[{"data":1,"prerenderedAt":1522},["ShallowReactive",2],{"portfolio-list-en":3},[4,321,556,843,1061,1279],{"id":5,"title":6,"body":7,"category":278,"client":279,"description":280,"displayTitle":281,"duration":286,"extension":287,"featured":288,"field":289,"gallery":290,"image":292,"meta":296,"navigation":288,"order":300,"participation":301,"path":302,"resultUrl":303,"role":304,"scope":305,"seo":306,"shortDescription":307,"slug":308,"stem":309,"technologies":310,"year":319,"__hash__":320},"portfolio_en\u002Fen\u002F01-ai-chatbot-service.md","Personalized Diet\u002FHealth Management Coaching Chatbot",{"type":8,"value":9,"toc":269},"minimark",[10,14,18,23,26,42,46,81,85,136,140,248,252],[11,12,6],"h1",{"id":13},"personalized-diethealth-management-coaching-chatbot",[15,16,17],"p",{},"A service where 3 chatbot coaches with distinct characters collect behavioral data through conversations with users and provide personalized coaching dialogue based on that data.",[19,20,22],"h2",{"id":21},"background","Background",[15,24,25],{},"For people on diets or managing their nutrition, a \"personalized management coach\" is essential.",[27,28,29,33,36,39],"ul",{},[30,31,32],"li",{},"Many people need someone who can give customized advice based on their eating and exercise patterns",[30,34,35],{},"Providing coaching with specific figures - which nutrients are lacking, how to exercise, how to balance exercise and food intake for weight loss - is tremendously helpful",[30,37,38],{},"Beyond knowledge-based advice, users need someone who can provide motivational conversations through encouragement, praise, and sometimes firm criticism. \"Character chatbot coaches\" were an excellent solution",[30,40,41],{},"Additionally, we focused on providing graphs, tables, and charts to help users easily and conveniently check accurate figures for systematic diet\u002Fhealth management",[19,43,45],{"id":44},"project-description","Project Description",[47,48,49,52,55,75,78],"ol",{},[30,50,51],{},"Three chatbot coaches with distinct characters collect behavioral data through conversations with users and provide personalized coaching dialogue.",[30,53,54],{},"Users select their preferred coach and chat like on KakaoTalk about what they eat, weight changes, exercise routines, and health conditions, receiving appropriate responses from the coach.",[30,56,57,58],{},"Content and features provided by the 3 character coaches:",[27,59,60,63,66,69,72],{},[30,61,62],{},"Analysis and advice on nutrients, calories for each user's meals",[30,64,65],{},"Analysis and advice on exercise amount and methods",[30,67,68],{},"Meal\u002Fexercise planning reflecting user's target period\u002Fweight",[30,70,71],{},"Statistics to view current nutrition status and past records at a glance",[30,73,74],{},"Integration with Google Health, Apple Health, and Samsung Health apps",[30,76,77],{},"The chatbot played a crucial role in providing not just knowledge-based advice but motivational conversations through encouragement, praise, and firm criticism.",[30,79,80],{},"With 3 chatbot characters having different personalities and conversation styles, users could enjoy finding a conversation style that suits them.",[19,82,84],{"id":83},"key-achievements","Key Achievements",[27,86,87,95,102,109,116,123,129],{},[30,88,89,90,94],{},"After commercialization, cumulative app downloads exceeded ",[91,92,93],"strong",{},"1.2 million"," (Android Google Play + iOS App Store combined)",[30,96,97,98,101],{},"Won ",[91,99,100],{},"\"Google Play Best Award\""," directly selected and announced by Google annually",[30,103,104,105,108],{},"Selected multiple times as ",[91,106,107],{},"\"Featured App\""," on Apple App Store and Google Play",[30,110,111,112,115],{},"Consistently maintained ",[91,113,114],{},"Top 10 ranking"," in Apple App Store's \"Health & Fitness\" category",[30,117,118,119,122],{},"User retention rate of ",[91,120,121],{},"35%",", about 3 times higher than average health apps (typical health app 2-month retention: 10-14%)",[30,124,125,128],{},[91,126,127],{},"57%"," of overweight users changed to normal weight range within 4 weeks (BMI basis)",[30,130,131,132,135],{},"Nutrient imbalance consumption rate ",[91,133,134],{},"decreased by over 32%"," after 50 days of app usage",[19,137,139],{"id":138},"development-process","Development Process",[47,141,142,158,174,190,206,222,235],{},[30,143,144,147],{},[91,145,146],{},"Step 1: Service Requirements Definition",[27,148,149,152,155],{},[30,150,151],{},"Identified user needs for health management",[30,153,154],{},"Defined key health management metrics important to users",[30,156,157],{},"Established \"character chatbot\" concept for \"personalized health coaching\"",[30,159,160,163],{},[91,161,162],{},"Step 2: User Scenario Confirmation",[27,164,165,168,171],{},[30,166,167],{},"Organized coaching content for health management",[30,169,170],{},"Prepared content for motivation: criticism, comfort, praise",[30,172,173],{},"Defined scenarios for recording and checking meals, exercise, weight",[30,175,176,179],{},[91,177,178],{},"Step 3: User Interface and App UX Design",[27,180,181,184,187],{},[30,182,183],{},"Designed interface for easy meal and exercise input",[30,185,186],{},"Built character traits and designed chatbot assets",[30,188,189],{},"Designed overall app structure and interface",[30,191,192,195],{},[91,193,194],{},"Step 4: Architecture Design, Development Planning",[27,196,197,200,203],{},[30,198,199],{},"Designed overall backend system and app structure",[30,201,202],{},"Designed and built database structure",[30,204,205],{},"Designed API structure using Event-sourcing, GRPC, GraphQL",[30,207,208,211],{},[91,209,210],{},"Step 5: Front-End (App), Backend Development",[27,212,213,216,219],{},[30,214,215],{},"Flutter-based app development",[30,217,218],{},"Kotlin, SpringBoot-based backend development",[30,220,221],{},"Built user analytics and push messaging systems with Mixpanel, Onesignal",[30,223,224,227],{},[91,225,226],{},"Step 6: Commercialization and User Feedback Collection",[27,228,229,232],{},[30,230,231],{},"Released on Google Play, iOS App Store",[30,233,234],{},"Conducted regular user interviews and behavioral data analysis",[30,236,237,240],{},[91,238,239],{},"Step 7: Chatbot Updates Based on User Feedback",[27,241,242,245],{},[30,243,244],{},"Identified update needs based on collected feedback",[30,246,247],{},"Improved user satisfaction\u002Fretention through regular updates",[19,249,251],{"id":250},"our-strengths","Our Strengths",[27,253,254,257,260,263,266],{},[30,255,256],{},"Development company with 10+ years of experience",[30,258,259],{},"Experience developing and operating services with 1.2+ million cumulative downloads",[30,261,262],{},"Google Play Best Award and Apple App Store\u002FGoogle Play Featured selection experience",[30,264,265],{},"Data\u002Falgorithm experts (Seoul National University ECE Bachelor's\u002FMaster's graduates)",[30,267,268],{},"Wishket Top 0.1% PRIME Partner certified",{"title":270,"searchDepth":271,"depth":271,"links":272},"",2,[273,274,275,276,277],{"id":21,"depth":271,"text":22},{"id":44,"depth":271,"text":45},{"id":83,"depth":271,"text":84},{"id":138,"depth":271,"text":139},{"id":250,"depth":271,"text":251},"Android, iOS","IT Startup (Series A+, TIPS Selected)","An AI healthcare service where 3 chatbot coaches with distinct characters collect behavioral data through conversations with users and provide personalized coaching based on that data",[282,283,284,285],"AI-Powered Personalized","Healthcare Platform,","Individual Health","Management Service","Jun 2023 - Mar 2024 (9 months)","md",true,"AI Healthcare",[291,292,293,294,295],"\u002Fimages\u002Fportfolio\u002F1\u002F1.png","\u002Fimages\u002Fportfolio\u002F1\u002F2.png","\u002Fimages\u002Fportfolio\u002F1\u002F3.png","\u002Fimages\u002Fportfolio\u002F1\u002F4.png","\u002Fimages\u002Fportfolio\u002F1\u002F5.png",{"ogImage":297,"sitemap":298},{"url":292},{"lastmod":299},"2024-03-15",1,"100%","\u002Fen\u002F01-ai-chatbot-service","https:\u002F\u002Fbalancefriends.com","Lead Product Design","Development, Design, Planning",{"title":6,"description":280},"iOS \u002F Android \u002F Kiosk","ai-chatbot-service","en\u002F01-ai-chatbot-service",[311,312,313,314,315,316,317,318],"Flutter","SpringBoot","Kotlin","MongoDB","MySQL","AWS","Google Gemini","DialogFlow","2023-2024","LDAqlbuVpxgrgCOl5EMnipc_xm-L5LLkTV5kxoN16e8",{"id":322,"title":323,"body":324,"category":521,"client":522,"description":523,"displayTitle":524,"duration":529,"extension":287,"featured":288,"field":530,"gallery":531,"image":533,"meta":537,"navigation":288,"order":271,"participation":301,"path":541,"resultUrl":542,"role":543,"scope":305,"seo":544,"shortDescription":545,"slug":546,"stem":547,"technologies":548,"year":554,"__hash__":555},"portfolio_en\u002Fen\u002F02-opencv-web-editor.md","AI Image Search, Generation\u002FEditing Editor",{"type":8,"value":325,"toc":514},[326,329,332,334,348,350,382,384,401,403,497,499],[11,327,323],{"id":328},"ai-image-search-generationediting-editor",[15,330,331],{},"A web editor that precisely detects object contours in images to automatically generate cut lines and provides AI-based image generation capabilities.",[19,333,22],{"id":21},[47,335,336,339,342,345],{},[30,337,338],{},"Image editing work in merchandise production is a time-consuming and labor-intensive process. Particularly for creating cut lines, client designers were manually doing this work. Additionally, inquiries from buyers seeking copyright-free high-resolution image sources were increasing.",[30,340,341],{},"To provide buyers with desired high-resolution images, reduce design work time, and minimize differences between requested designs and actual products, an editor utilizing generative AI was needed where buyers could complete designs themselves.",[30,343,344],{},"An intuitive interface was needed that not only designers but also general users could easily use, with AI technology automating image generation and processing to reduce user workload.",[30,346,347],{},"This project aimed to go beyond simple image editing tools to maximize interaction between users and generative AI to support creative work.",[19,349,45],{"id":44},[47,351,352,358,364,370,376],{},[30,353,354,357],{},[91,355,356],{},"Contour Detection Algorithm",": Developed an algorithm that precisely detects object contours in images to automatically generate cut lines. Optimized to include all images even in complex forms like separated images or dotted images.",[30,359,360,363],{},[91,361,362],{},"AI-Based Image Generation",": Provided functionality where users enter keywords for desired images and AI generates 3 high-resolution image candidates to choose from.",[30,365,366,369],{},[91,367,368],{},"Image Editing Tools",": Provided intuitive editing tools enabling general users unfamiliar with design tools like Photoshop to freely modify and adjust images. Key features include color adjustment, text addition, background removal compositing, and resizing.",[30,371,372,375],{},[91,373,374],{},"Product-Specific Options",": Administrators can register various options by merchandise type (e.g., keyring hole size\u002Fposition, stand length\u002Fposition) in the editor, allowing users to directly edit options.",[30,377,378,381],{},[91,379,380],{},"Collaboration and Workflow Integration",": Designed APIs to facilitate collaboration with external development companies building the shopping mall.",[19,383,84],{"id":83},[27,385,386,389,392,395,398],{},[30,387,388],{},"Buyers can complete edits themselves, reducing internal designer resource allocation",[30,390,391],{},"General users can create professional-level edits without specialized knowledge of tools like Photoshop",[30,393,394],{},"Customer inquiries about copyright and image provision decreased with generative AI assistance",[30,396,397],{},"Time from design to final confirmation significantly reduced",[30,399,400],{},"Actual buyer surveys and reviews showed high satisfaction with editor quality, speed, and AI image feature convenience",[19,402,139],{"id":138},[47,404,405,417,429,442,455,468,481],{},[30,406,407,409],{},[91,408,146],{},[27,410,411,414],{},[30,412,413],{},"Researched various image editing tools",[30,415,416],{},"Established detailed requirements based on client customer inquiries",[30,418,419,421],{},[91,420,162],{},[27,422,423,426],{},[30,424,425],{},"Prioritized features by examining actual design samples sent to print shops",[30,427,428],{},"Confirmed user editor usage flow",[30,430,431,434],{},[91,432,433],{},"Step 3: User Interface and Editor UX Design",[27,435,436,439],{},[30,437,438],{},"Designed intuitive interface for general users unfamiliar with design",[30,440,441],{},"Designed UX not deviating significantly from popular editing tool conventions",[30,443,444,447],{},[91,445,446],{},"Step 4: Contour Detection and Cut Line Generation Algorithm Design",[27,448,449,452],{},[30,450,451],{},"Designed to detect contours without missing any in images with 2+ parts or dotted patterns",[30,453,454],{},"Designed adjustable distance between contours and cut lines",[30,456,457,460],{},[91,458,459],{},"Step 5: Editor Feature Development",[27,461,462,465],{},[30,463,464],{},"Developed basic editor features: photo upload, text insertion\u002Fediting, image rotation",[30,466,467],{},"Developed using JavaScript, Vue.js, TypeScript",[30,469,470,473],{},[91,471,472],{},"Step 6: Testing and Tester Feedback Collection",[27,474,475,478],{},[30,476,477],{},"Compared files completed by buyers vs. files processed by designers",[30,479,480],{},"Analyzed user editor usage behavior and conducted interviews",[30,482,483,486],{},[91,484,485],{},"Step 7: Feedback-Based Updates and Commercialization",[27,487,488,491,494],{},[30,489,490],{},"Modified UI\u002FUX based on collected feedback",[30,492,493],{},"Reduced cut line generation time",[30,495,496],{},"Improved cut line generation algorithm logic",[19,498,251],{"id":250},[27,500,501,503,505,508,511],{},[30,502,256],{},[30,504,265],{},[30,506,507],{},"Proprietary solution for cut line generation through contour detection",[30,509,510],{},"Image and text editor development and operation experience",[30,512,513],{},"Shopping mall development experience with strong collaboration and communication capabilities",{"title":270,"searchDepth":271,"depth":271,"links":515},[516,517,518,519,520],{"id":21,"depth":271,"text":22},{"id":44,"depth":271,"text":45},{"id":83,"depth":271,"text":84},{"id":138,"depth":271,"text":139},{"id":250,"depth":271,"text":251},"Web (PC\u002FMobile)","IT Startup","A web editor that precisely detects contours in images to automatically generate cut lines and provides AI-based image generation capabilities",[525,526,527,528],"Machine Learning-Based","Image Processing Technology,","OpenCV-Powered","Web Editor","May 2024 - Jul 2024 (2 months)","AI Image Processing",[532,533,534,535,536],"\u002Fimages\u002Fportfolio\u002F2\u002F1.png","\u002Fimages\u002Fportfolio\u002F2\u002F2.png","\u002Fimages\u002Fportfolio\u002F2\u002F3.png","\u002Fimages\u002Fportfolio\u002F2\u002F4.png","\u002Fimages\u002Fportfolio\u002F2\u002F5.png",{"ogImage":538,"sitemap":539},{"url":533},{"lastmod":540},"2024-07-15","\u002Fen\u002F02-opencv-web-editor",null,"Algorithm & UI\u002FUX Design",{"title":323,"description":523},"Vue.js \u002F TypeScript \u002F OpenCV","opencv-web-editor","en\u002F02-opencv-web-editor",[549,550,551,552,553],"JavaScript","Vue.js","TypeScript","OpenAI","DALL·E 2","2024","NJj-JXscCnM_HWeNKKgRgvysySiW9jdLYkpjwo_mWB0",{"id":557,"title":558,"body":559,"category":521,"client":810,"description":811,"displayTitle":812,"duration":817,"extension":287,"featured":288,"field":818,"gallery":819,"image":820,"meta":825,"navigation":288,"order":829,"participation":301,"path":830,"resultUrl":542,"role":831,"scope":305,"seo":832,"shortDescription":833,"slug":834,"stem":835,"technologies":836,"year":554,"__hash__":842},"portfolio_en\u002Fen\u002F03-ai-customer-center.md","Customer Inquiry Analysis and Response Generation Automation: AI Customer Center",{"type":8,"value":560,"toc":803},[561,564,567,569,583,585,659,661,691,693,786,788],[11,562,558],{"id":563},"customer-inquiry-analysis-and-response-generation-automation-ai-customer-center",[15,565,566],{},"A system that builds a chatbot-style customer center to efficiently handle FAQs and option\u002Fchat-type inquiries.",[19,568,22],{"id":21},[27,570,571,574,577,580],{},[30,572,573],{},"The client operating a large shopping mall had at least 7 customer service agents directly responding to various customer inquiries daily. On high-volume days or when complex inquiries extended consultation time, customer satisfaction was significantly declining.",[30,575,576],{},"Although FAQs and inquiry boards were utilized, users preferred chat-based consultations. However, the existing KakaoTalk channel-based option selection customer center had relatively low customer satisfaction.",[30,578,579],{},"The client was continuously expanding the shopping mall, meaning customer inquiries would increase further, requiring an effective solution.",[30,581,582],{},"Hiring additional customer service staff for temporarily increasing inquiries was burdensome, and utilizing existing staff for other tasks would be more effective.",[19,584,45],{"id":44},[47,586,587,600,610,623,636,649],{},[30,588,589,592],{},[91,590,591],{},"Past Customer Inquiry Analysis and Key Scenario Identification",[27,593,594,597],{},[30,595,596],{},"Categorized FAQs and situations requiring\u002Fnot requiring chat",[30,598,599],{},"Organized and classified newly needed response situations",[30,601,602,605],{},[91,603,604],{},"User Interface Design (Including Design)",[27,606,607],{},[30,608,609],{},"Another key aspect of AI chatbot projects is \"efficiently exchanging Input\u002FOutput within a defined chat space\" - performed scenario analysis and design work for this",[30,611,612,615],{},[91,613,614],{},"Backend System Design and Development",[27,616,617,620],{},[30,618,619],{},"Determined which generative AI solution to use and when\u002Fhow to call APIs",[30,621,622],{},"Comprehensive design and implementation of how to connect with shopping mall member\u002Fproduct databases",[30,624,625,628],{},[91,626,627],{},"Question-Answer Database Construction",[27,629,630,633],{},[30,631,632],{},"Built Q&A DB for FAQs, option\u002Fchat-type inquiries classified earlier",[30,634,635],{},"Determined and executed cost-efficient service operations",[30,637,638,641],{},[91,639,640],{},"Real Agent and AI Chatbot Collaboration Scenario Design and Implementation",[27,642,643,646],{},[30,644,645],{},"Even with AI chatbots, real agent intervention is needed in certain exception situations",[30,647,648],{},"Implemented features for real agents to collaborate with AI chatbot for quality management of early responses",[30,650,651,654],{},[91,652,653],{},"Continuous Learning Updates",[27,655,656],{},[30,657,658],{},"Performed ongoing Prompt Engineering and Fine Tuning for response optimization",[19,660,84],{"id":83},[27,662,663,670,681,688],{},[30,664,665,666,669],{},"Customer inquiries that were only handled during agent working hours (Mon-Fri, 10AM-5PM) can now be processed ",[91,667,668],{},"regardless of day\u002Ftime",", greatly improving user satisfaction",[30,671,672,673,676],{},"With only a few given options like \"Order inquiry,\" \"Shipping policy,\" \"Payment methods,\" the probability of customers finding desired answers was very low. The chatbot-style customer center achieved ",[91,674,675],{},"80-90% \"first response\" completion rate",[27,677,678],{},[30,679,680],{},"(10-20% of inquiries were directly handled by agents while strengthening chatbot capabilities)",[30,682,683,684,687],{},"Handled more inquiries with only ",[91,685,686],{},"20-30% of the human resources"," previously allocated for customer service",[30,689,690],{},"Continuously updated Vector DB (RAG) after commercialization to improve response rates",[19,692,139],{"id":138},[47,694,695,710,725,741,757,773],{},[30,696,697,699],{},[91,698,146],{},[27,700,701,704,707],{},[30,702,703],{},"Established detailed customer center requirements (including AI response and agent collaboration features)",[30,705,706],{},"Collected sales policies, shipping policies, FAQs",[30,708,709],{},"Organized and cleaned existing inquiry\u002Fresponse records",[30,711,712,714],{},[91,713,162],{},[27,715,716,719,722],{},[30,717,718],{},"Confirmed chatbot response scenarios (flow)",[30,720,721],{},"Basic Prompt Engineering design",[30,723,724],{},"Confirmed chatbot response feedback verification and human agent collaboration methods",[30,726,727,730],{},[91,728,729],{},"Step 3: User Interface and Web UX Design",[27,731,732,735,738],{},[30,733,734],{},"Reviewed and adjusted service requirements and user scenarios",[30,736,737],{},"Determined User Interface capable of handling scenarios",[30,739,740],{},"Designed Web Application features",[30,742,743,746],{},[91,744,745],{},"Step 4: Architecture Design, AI Solution Integration",[27,747,748,751,754],{},[30,749,750],{},"Designed backend system and database configuration",[30,752,753],{},"Built RAG system using VectorDB (Weaviate)",[30,755,756],{},"Integrated LLMs including OpenAI, Gemini",[30,758,759,762],{},[91,760,761],{},"Step 5: Full Feature Development and Commercialization",[27,763,764,767,770],{},[30,765,766],{},"Front-End development",[30,768,769],{},"Back-End development",[30,771,772],{},"Collected user feedback through commercialization",[30,774,775,778],{},[91,776,777],{},"Step 6: Continuous Tuning (AI)",[27,779,780,783],{},[30,781,782],{},"Prompt Engineering to improve response rates",[30,784,785],{},"Updates based on user feedback analysis",[19,787,251],{"id":250},[27,789,790,792,794,797,800],{},[30,791,256],{},[30,793,265],{},[30,795,796],{},"Rich experience planning\u002Fdesigning\u002Fdeveloping\u002Fcommercializing \"conversational chatbot\" service with 1.2+ million downloads",[30,798,799],{},"Proven capability in user analysis and excellent UX\u002FUI, as demonstrated by Google Play Best Award (selected once per year by Google)",[30,801,802],{},"Extensive project experience in automation\u002Fcontent generation using generative AI solutions",{"title":270,"searchDepth":271,"depth":271,"links":804},[805,806,807,808,809],{"id":21,"depth":271,"text":22},{"id":44,"depth":271,"text":45},{"id":83,"depth":271,"text":84},{"id":138,"depth":271,"text":139},{"id":250,"depth":271,"text":251},"Mid-sized Company (Industry #2, Revenue 500B+ KRW)","A system that builds a chatbot-style customer center to efficiently handle FAQs and option\u002Fchat-type inquiries, implementing collaboration scenarios between real agents and AI chatbots",[813,814,815,816],"Intelligent AI","Customer Response System,","VectorDB-Based","Customer Satisfaction Solution","Oct 2024 - Feb 2025 (5 months)","AI Customer Service",[820,821,822,823,824],"\u002Fimages\u002Fportfolio\u002F3\u002F1.png","\u002Fimages\u002Fportfolio\u002F3\u002F2.png","\u002Fimages\u002Fportfolio\u002F3\u002F3.png","\u002Fimages\u002Fportfolio\u002F3\u002F4.png","\u002Fimages\u002Fportfolio\u002F3\u002F5.png",{"ogImage":826,"sitemap":827},{"url":820},{"lastmod":828},"2025-02-15",3,"\u002Fen\u002F03-ai-customer-center","System Design & AI Integration",{"title":558,"description":811},"Node.js \u002F MongoDB \u002F VectorDB","ai-customer-center","en\u002F03-ai-customer-center",[314,837,311,838,839,313,840,552,841],"VectorDB (Weaviate)","Node.js","Redis","RabbitMQ","Gemini","6fQ68NWdvEY5VPtQdH4M8t6yvRv5u3MSz4hq-PIt7o4",{"id":844,"title":845,"body":846,"category":278,"client":279,"description":1032,"displayTitle":1033,"duration":1038,"extension":287,"featured":1039,"field":1040,"gallery":1041,"image":1042,"meta":1047,"navigation":288,"order":1051,"participation":301,"path":1052,"resultUrl":542,"role":1053,"scope":305,"seo":1054,"shortDescription":1055,"slug":1056,"stem":1057,"technologies":1058,"year":554,"__hash__":1060},"portfolio_en\u002Fen\u002F04-ai-image-analysis.md","AI Image Analysis and Information Extraction, Data Entry Automation",{"type":8,"value":847,"toc":1025},[848,851,854,856,870,872,886,888,913,915,1012,1014],[11,849,845],{"id":850},"ai-image-analysis-and-information-extraction-data-entry-automation",[15,852,853],{},"A project that automates the repetitive task of recognizing photo (image) format input data, extracting specific text and corresponding numerical values, and converting them into database entries.",[19,855,22],{"id":21},[27,857,858,861,864,867],{},[30,859,860],{},"Previously, at least 1.5 Man\u002FMonth was required for dedicated personnel to organize characters and numbers from similarly formatted tables into Excel files",[30,862,863],{},"While table formats and sizes were similar, i) they weren't perfectly identical, and ii) key value order was inconsistent, requiring significant concentration when performed manually",[30,865,866],{},"Since the task involved recognizing image-format materials and entering them into Excel, human errors were inevitable, and considering additional resources needed to find\u002Fcorrect such errors, End-to-End task required 2+ Man\u002FMonth",[30,868,869],{},"To reduce 2+ Man\u002FMonth resources and lower error probability, \"AI-based image analysis and entry automation\" was essential",[19,871,45],{"id":44},[27,873,874,877,880,883],{},[30,875,876],{},"Users submit photos of table-format images containing text (Key values) and numbers (Value values), which are analyzed using AI solutions",[30,878,879],{},"Recognizes text to find matching Key values and automatically enters corresponding Value values, saving them to the database automatically",[30,881,882],{},"Even if text (Key values)-numbers (Value values) in user-submitted photos are in inconsistent order, Values are automatically sorted to match Keys",[30,884,885],{},"Since images are input via photography, \"image analysis tuning work\" for distinguishing \"3\" from \"8\" or \"number 9\" from \"letter g\" was crucial",[19,887,84],{"id":83},[27,889,890,897,904,910],{},[30,891,892,893,896],{},"Eliminated the need for ",[91,894,895],{},"2+ Man\u002FMonth human resource investment"," previously required for manual data entry",[30,898,899,900,903],{},"Reduced error probability from mistakes or typos from ~1-2% to ",[91,901,902],{},"0.1-0.2%",", a 1\u002F10 reduction, also improving efficiency of subsequent data-utilizing work",[30,905,906,907],{},"When humans manually entered data, entering and reviewing 40-50 data points took about 4-5 hours; with AI, data entry time ",[91,908,909],{},"decreased dramatically to under 10 minutes",[30,911,912],{},"Using images (photography) also simplified the data entry process",[19,914,139],{"id":138},[47,916,917,929,944,959,973,987,999],{},[30,918,919,921],{},[91,920,146],{},[27,922,923,926],{},[30,924,925],{},"Confirmed basic format of images to be input",[30,927,928],{},"Classified various image examples\u002Fformats and performed initial data cleaning",[30,930,931,933],{},[91,932,162],{},[27,934,935,938,941],{},[30,936,937],{},"Confirmed user image input scenarios",[30,939,940],{},"Established flow for photo capture, image confirmation, input value verification",[30,942,943],{},"Organized methods to request accurate image capture from users",[30,945,946,948],{},[91,947,178],{},[27,949,950,953,956],{},[30,951,952],{},"Developed User Interface for easy image input and data verification",[30,954,955],{},"Designed overall App UX including additional features",[30,957,958],{},"Developed Interface to provide detailed image capture\u002Finput guidance to users",[30,960,961,963],{},[91,962,745],{},[27,964,965,967,970],{},[30,966,750],{},[30,968,969],{},"Developed image recognition features using Gemini",[30,971,972],{},"Performed Fine-Tuning",[30,974,975,977],{},[91,976,761],{},[27,978,979,982,985],{},[30,980,981],{},"Front-End development (Flutter, etc.)",[30,983,984],{},"Back-End development (Node.js, etc.)",[30,986,772],{},[30,988,989,992],{},[91,990,991],{},"Step 6: Continuous AI Tuning",[27,993,994,997],{},[30,995,996],{},"Gemini-related tuning to improve image recognition rates",[30,998,785],{},[30,1000,1001,1004],{},[91,1002,1003],{},"Step 7: Internal Review and Comparison Tuning",[27,1005,1006,1009],{},[30,1007,1008],{},"Fine-Tuning through comparison of internal review results and actual recognition results",[30,1010,1011],{},"Strengthened user communication for better image quality",[19,1013,251],{"id":250},[27,1015,1016,1018,1020,1022],{},[30,1017,256],{},[30,1019,265],{},[30,1021,268],{},[30,1023,1024],{},"Experience with multiple AI solutions and AI-based image editor development",{"title":270,"searchDepth":271,"depth":271,"links":1026},[1027,1028,1029,1030,1031],{"id":21,"depth":271,"text":22},{"id":44,"depth":271,"text":45},{"id":83,"depth":271,"text":84},{"id":138,"depth":271,"text":139},{"id":250,"depth":271,"text":251},"An AI solution that recognizes photo (image) format input data, extracts text and numerical values, and automates the process of converting them into database entries",[1034,1035,1036,1037],"AI Image Recognition","Automation Solution,","Data Extraction and","Entry Automation","Feb 2024 - May 2024 (3 months)",false,"AI Data Automation",[1042,1043,1044,1045,1046],"\u002Fimages\u002Fportfolio\u002F5\u002F1.png","\u002Fimages\u002Fportfolio\u002F5\u002F2.png","\u002Fimages\u002Fportfolio\u002F5\u002F3.png","\u002Fimages\u002Fportfolio\u002F5\u002F기능 03__AI기반 자동 입력 기능.png","\u002Fimages\u002Fportfolio\u002F5\u002F기능 04_오류 검출 및 자동 보정.png",{"ogImage":1048,"sitemap":1049},{"url":1042},{"lastmod":1050},"2024-05-15",4,"\u002Fen\u002F04-ai-image-analysis","AI Solution Design & Development",{"title":845,"description":1032},"Flutter \u002F Google Gemini \u002F MongoDB","ai-image-analysis","en\u002F04-ai-image-analysis",[311,312,313,314,315,1059,317],"Google Cloud","KQkcb2SN9-hiwfztL_qqK9a8hDGJrnTOLvUY2nSuRdw",{"id":1062,"title":1063,"body":1064,"category":278,"client":279,"description":1252,"displayTitle":1253,"duration":1258,"extension":287,"featured":1039,"field":1259,"gallery":1260,"image":1261,"meta":1266,"navigation":288,"order":1270,"participation":301,"path":1271,"resultUrl":542,"role":1272,"scope":305,"seo":1273,"shortDescription":1274,"slug":1275,"stem":1276,"technologies":1277,"year":319,"__hash__":1278},"portfolio_en\u002Fen\u002F05-ai-product-recommendation.md","Personalized Product Recommendation and Purchase Conversion",{"type":8,"value":1065,"toc":1245},[1066,1069,1072,1074,1088,1090,1107,1109,1112,1134,1137,1152,1154,1229,1231],[11,1067,1063],{"id":1068},"personalized-product-recommendation-and-purchase-conversion",[15,1070,1071],{},"An AI solution that analyzes diverse user preferences and requirements to recommend personalized products for a shopping mall selling various food items targeting women in their 20s-30s.",[19,1073,22],{"id":21},[47,1075,1076,1079,1082,1085],{},[30,1077,1078],{},"Finding suitable products in a shopping mall is always a tedious and time-consuming task for users. The more product variety, the greater the inconvenience, requiring a solution.",[30,1080,1081],{},"Users have different criteria for selecting products, considering both quantitative criteria (performance-based) and qualitative criteria (taste and preferences), adding complexity to recommendation algorithms.",[30,1083,1084],{},"By learning quantitative and qualitative criteria and deriving products from candidate groups that best satisfy these criteria, AI could significantly improve recommendation accuracy.",[30,1086,1087],{},"Since users could input their preference information in certain situations, we could progressively improve recommendation accuracy.",[19,1089,45],{"id":44},[47,1091,1092,1095,1098,1101,1104],{},[30,1093,1094],{},"Planned\u002Fdesigned\u002Fdeveloped\u002Fcommercialized for a shopping mall selling various food items to women in their 20s-30s",[30,1096,1097],{},"With women in their 20s-30s as main customers, requirements for taste\u002Fpreferences\u002Fpackaging\u002Fprice range varied greatly, making accurate product recommendations crucial",[30,1099,1100],{},"For some customers, bundling multiple products into packages with slight discounts increased purchase conversion rates. AI was also utilized for personalized package configuration.",[30,1102,1103],{},"To understand customer preferences, we acquired information needed for recommendations through simple Q&A conversations. Planning and design for efficient recommendation questions and priority criteria when customer requirements conflict was also crucial.",[30,1105,1106],{},"High user satisfaction was achieved by considering both i) quantitative information and criteria and ii) qualitative characteristics and tendencies in personalized recommendations.",[19,1108,84],{"id":83},[15,1110,1111],{},"After applying personalized recommendations:",[27,1113,1114,1119,1124,1129],{},[30,1115,1116],{},[91,1117,1118],{},"Product page view rate increased approximately 10x",[30,1120,1121],{},[91,1122,1123],{},"Purchase process entry rate increased approximately 7x",[30,1125,1126],{},[91,1127,1128],{},"Purchase success rate increased approximately 5x",[30,1130,1131],{},[91,1132,1133],{},"Total revenue also increased approximately 5x",[15,1135,1136],{},"Additional achievements:",[27,1138,1139,1146],{},[30,1140,1141,1142,1145],{},"Personalized recommendations enabled introducing more products to users, ",[91,1143,1144],{},"increasing displayable product area by 3-4x",", removing burden from continuously expanding product lines",[30,1147,1148,1149],{},"With \"Algorithm\u002FRule-based\" approaches, when i) factors to consider for recommendations change or ii) quantitative\u002Fqualitative criteria values change, the algorithm itself needed modification, requiring separate resource allocation. With ",[91,1150,1151],{},"AI solutions, updates could be done simply and easily",[19,1153,139],{"id":138},[47,1155,1156,1168,1180,1193,1205,1216],{},[30,1157,1158,1160],{},[91,1159,146],{},[27,1161,1162,1165],{},[30,1163,1164],{},"Designed user analysis and preference data collection methods",[30,1166,1167],{},"Defined quantitative\u002Fqualitative criteria for recommendation algorithms",[30,1169,1170,1172],{},[91,1171,162],{},[27,1173,1174,1177],{},[30,1175,1176],{},"Designed preference identification scenarios through Q&A conversations",[30,1178,1179],{},"Established recommendation result display and feedback collection flow",[30,1181,1182,1185],{},[91,1183,1184],{},"Step 3: Recommendation Algorithm Design",[27,1186,1187,1190],{},[30,1188,1189],{},"Integrated quantitative criteria (price, nutrients) with qualitative criteria (taste, preferences)",[30,1191,1192],{},"Designed data structure for AI model training",[30,1194,1195,1198],{},[91,1196,1197],{},"Step 4: Architecture Design and Development",[27,1199,1200,1202],{},[30,1201,750],{},[30,1203,1204],{},"AI recommendation engine integration and API development",[30,1206,1207,1210],{},[91,1208,1209],{},"Step 5: Front-End, BackEnd Development",[27,1211,1212,1214],{},[30,1213,215],{},[30,1215,218],{},[30,1217,1218,1221],{},[91,1219,1220],{},"Step 6: Commercialization and Performance Measurement",[27,1222,1223,1226],{},[30,1224,1225],{},"Validated recommendation effectiveness through A\u002FB testing",[30,1227,1228],{},"Continuous improvement based on user feedback",[19,1230,251],{"id":250},[27,1232,1233,1235,1238,1241,1243],{},[30,1234,256],{},[30,1236,1237],{},"Experience applying \"content recommendation algorithms\" to actual services achieving 100K+ downloads",[30,1239,1240],{},"4 patents in \"Big Data-Based Personalized Recommendation Algorithm\" field (filing in progress)",[30,1242,265],{},[30,1244,268],{},{"title":270,"searchDepth":271,"depth":271,"links":1246},[1247,1248,1249,1250,1251],{"id":21,"depth":271,"text":22},{"id":44,"depth":271,"text":45},{"id":83,"depth":271,"text":84},{"id":138,"depth":271,"text":139},{"id":250,"depth":271,"text":251},"An AI solution that analyzes user preferences to provide personalized product recommendations and package configurations, optimizing shopping malls with AI-based recommendation algorithms",[1254,1255,1256,1257],"AI-Based","Personalized Product Recommendation,","Purchase Conversion","Maximization Solution","Jun 2023 - Feb 2024 (8 months)","AI Recommendation System",[1261,1262,1263,1264,1265],"\u002Fimages\u002Fportfolio\u002F6\u002F1.png","\u002Fimages\u002Fportfolio\u002F6\u002F2.png","\u002Fimages\u002Fportfolio\u002F6\u002F3.png","\u002Fimages\u002Fportfolio\u002F6\u002F기능 03_사용자 피드백 반영 시스템.png","\u002Fimages\u002Fportfolio\u002F6\u002F기능 04_유연한 추천 기준 설정 및 변경.png",{"ogImage":1267,"sitemap":1268},{"url":1261},{"lastmod":1269},"2024-02-15",5,"\u002Fen\u002F05-ai-product-recommendation","Algorithm Design & System Architecture",{"title":1063,"description":1252},"Flutter \u002F AI Recommendation \u002F E-commerce","ai-product-recommendation","en\u002F05-ai-product-recommendation",[311,312,313,314,315,1059,316,552,841],"nVk20FEJT0K0Jxf_CtVmaRWcoi3m85EfdS38cPBnrT0",{"id":1280,"title":1281,"body":1282,"category":278,"client":279,"description":1496,"displayTitle":1497,"duration":1501,"extension":287,"featured":1039,"field":1502,"gallery":1503,"image":1504,"meta":1509,"navigation":288,"order":1512,"participation":301,"path":1513,"resultUrl":542,"role":1514,"scope":305,"seo":1515,"shortDescription":1516,"slug":1517,"stem":1518,"technologies":1519,"year":319,"__hash__":1521},"portfolio_en\u002Fen\u002F06-ai-report-automation.md","Report Generation Automation",{"type":8,"value":1283,"toc":1489},[1284,1287,1290,1292,1320,1322,1325,1333,1336,1362,1364,1389,1391,1473,1475],[11,1285,1281],{"id":1286},"report-generation-automation",[15,1288,1289],{},"A solution that automatically generates comprehensive reports using AI to analyze user data entered periodically, including graphing, evaluation, prediction, and text expression.",[19,1291,22],{"id":21},[47,1293,1294,1314,1317],{},[30,1295,1296,1297],{},"Designing\u002Fdeveloping a solution that performs all the following is not straightforward:",[27,1298,1299,1302,1305,1308,1311],{},[30,1300,1301],{},"Algorithm\u002Fsolution for expressing data as graphs",[30,1303,1304],{},"Algorithm\u002Fsolution for comparing and analyzing past and present data",[30,1306,1307],{},"Algorithm\u002Fsolution for making comprehensive evaluations based on data",[30,1309,1310],{},"Algorithm\u002Fsolution for predicting future values from data analysis",[30,1312,1313],{},"Algorithm\u002Fsolution for expressing data analysis results in text",[30,1315,1316],{},"We determined these features could be developed faster and easier with better performance using generative AI including LLM solutions",[30,1318,1319],{},"We judged cost-efficient implementation and results could be achieved with appropriate Prompt Engineering and caching",[19,1321,45],{"id":44},[15,1323,1324],{},"Users periodically entered personal data for several items:",[27,1326,1327,1330],{},[30,1328,1329],{},"While the total types of items users could enter were defined, which items were entered varied by situation each time",[30,1331,1332],{},"Continuous and discontinuous data were mixed",[15,1334,1335],{},"In this situation:",[47,1337,1338,1344,1350,1356],{},[30,1339,1340,1343],{},[91,1341,1342],{},"Data Analysis",": Analyzed data entry presence, entered data values, data changes, and trends",[30,1345,1346,1349],{},[91,1347,1348],{},"Evaluation and Scoring",": Made judgments (Good or Bad) about users based on analysis results and expressed the degree as numerical scores",[30,1351,1352,1355],{},[91,1353,1354],{},"Text Expression",": Beyond expressing analysis scores, presented data analysis content in user-friendly wording for easy understanding",[30,1357,1358,1361],{},[91,1359,1360],{},"Prediction Data",": Provided not only data-based analysis results but also future prediction data",[19,1363,84],{"id":83},[27,1365,1366,1372,1382],{},[30,1367,1368,1369],{},"Completed all algorithms for expressing data as graphs, comparing past\u002Fpresent, comprehensively analyzing and evaluating, predicting future, and expressing results in text within 1 month with high quality. Considering 2-3 months development time if done separately, ",[91,1370,1371],{},"reduced development resources by approximately 50-60%",[30,1373,1374,1375,1378,1379],{},"Using generative AI, data analysis results could be expressed in ",[91,1376,1377],{},"\"easy to read, easy to understand sentence\""," format. Core summaries were also easily achievable. This improved readability and communication for users, ",[91,1380,1381],{},"significantly increasing data analysis report satisfaction",[30,1383,1384,1385,1388],{},"Could ",[91,1386,1387],{},"flexibly accommodate"," slight changes in report data types, content, and format and immediately reflect them in analysis. If features were implemented separately using traditional methods, additional development resources would have been needed to respond to such changes",[19,1390,139],{"id":138},[47,1392,1393,1405,1418,1434,1447,1460],{},[30,1394,1395,1397],{},[91,1396,146],{},[27,1398,1399,1402],{},[30,1400,1401],{},"Defined analysis target data types and formats",[30,1403,1404],{},"Confirmed analysis items to include in reports",[30,1406,1407,1410],{},[91,1408,1409],{},"Step 2: Data Analysis Design",[27,1411,1412,1415],{},[30,1413,1414],{},"Designed continuous\u002Fdiscontinuous data processing approaches",[30,1416,1417],{},"Defined evaluation criteria and scoring methods",[30,1419,1420,1423],{},[91,1421,1422],{},"Step 3: AI Solution Architecture Design",[27,1424,1425,1428,1431],{},[30,1426,1427],{},"Selected LLM models (Claude, OpenAI)",[30,1429,1430],{},"Established Prompt Engineering strategy",[30,1432,1433],{},"Cost optimization through caching strategy",[30,1435,1436,1439],{},[91,1437,1438],{},"Step 4: Visualization and Report Template Development",[27,1440,1441,1444],{},[30,1442,1443],{},"Developed graph\u002Fchart generation modules",[30,1445,1446],{},"Designed report layouts",[30,1448,1449,1452],{},[91,1450,1451],{},"Step 5: Backend System Development",[27,1453,1454,1457],{},[30,1455,1456],{},"Node.js, SpringBoot-based API development",[30,1458,1459],{},"MongoDB, VectorDB integration",[30,1461,1462,1465],{},[91,1463,1464],{},"Step 6: Commercialization and Feedback Collection",[27,1466,1467,1470],{},[30,1468,1469],{},"Real user testing and feedback collection",[30,1471,1472],{},"Report quality improvement",[19,1474,251],{"id":250},[27,1476,1477,1479,1481,1484,1486],{},[30,1478,256],{},[30,1480,265],{},[30,1482,1483],{},"Rich experience developing\u002Fcommercializing hospital\u002Fpatient data analysis services like 'PatientTree', operated as paid services in 100+ hospitals providing hospital-specific analysis data and reports",[30,1485,268],{},[30,1487,1488],{},"Extensive experience conducting data analysis using multiple AI solutions",{"title":270,"searchDepth":271,"depth":271,"links":1490},[1491,1492,1493,1494,1495],{"id":21,"depth":271,"text":22},{"id":44,"depth":271,"text":45},{"id":83,"depth":271,"text":84},{"id":138,"depth":271,"text":139},{"id":250,"depth":271,"text":251},"A comprehensive report automation solution utilizing generative AI for data analysis, visualization, evaluation, prediction, and text expression",[1254,1498,1499,1500],"Data Analysis Report,","Automatic Generation","Solution","Oct 2023 - Mar 2024 (4 months)","AI Data Analysis",[1504,1505,1506,1507,1508],"\u002Fimages\u002Fportfolio\u002F4\u002F1.png","\u002Fimages\u002Fportfolio\u002F4\u002F2.png","\u002Fimages\u002Fportfolio\u002F4\u002F3.png","\u002Fimages\u002Fportfolio\u002F4\u002F기능 03_사용자 맞춤형 분석 결과 제공.png","\u002Fimages\u002Fportfolio\u002F4\u002F기능 04_분석 기준의 유연한 설정.png",{"ogImage":1510,"sitemap":1511},{"url":1504},{"lastmod":299},6,"\u002Fen\u002F06-ai-report-automation","AI Solution Architecture & Development",{"title":1281,"description":1496},"Node.js \u002F Claude \u002F Data Analytics","ai-report-automation","en\u002F06-ai-report-automation",[838,312,313,314,837,1059,316,1520,552],"Claude","xA3JQIhpFUSwIQV8UkAlR0MurzGJ8qAVq-yrNMZsUJU",1789708273158]