Architected a multi-platform social ecosystem (Mobile/Flutter, Web/Next.js) powered by a microservices backend.

Integrated multi-modal LLMs (Gemini, OpenAI, Claude) to drive core features including automated "Life Story" generation, AI-moderated content feeds, and on-device image validation/transformation.

Engineered vector-based embedding systems for high-precision user matching and developed a custom Random Walker algorithm to cluster 30k+ geospatial "memories," creating a dynamic and diverse social map.

Built robust Python/Selenium scraping pipelines to aggregate and process social and event data, utilizing AI agents for "Deep Research" to enrich content and automate event storytelling.

Deployed serverless infrastructure on Vercel and Supabase with automated CI/CD workflows (GitHub Actions).

Developed a real-time analytics dashboard to track user growth and engagement metrics across the entire platform ecosystem.

Created a user-matching service using TypeScript and Supabase, utilizing cosine similarity for improved recommendation accuracy by 20%.

Established a performant, scalable, real-time search engine for finding friends using Next.js, Supabase, and Vercel KV.

Designed and implemented a chat-simulation API leveraging the OpenAI API for dynamic, AI-driven conversations, enhancing user engagement.

Optimized backend architecture for scalability, reducing response times by 30% and ensuring seamless integration with front-end services.

Utilized TypeScript and modular component design to maintain high code quality and scalability, supporting a 35% increase in concurrent users without performance issues.

Designed and set up comprehensive navigation in Next.js, achieving a 30% increase in user engagement through seamless page transitions.

Created dynamic camera and profile selection features, boosting user interaction with personalized media content by 20%.

Spearheaded user tracking with Firebase and Google Analytics, accelerating new user registration by 50%.

Analyzed performance trade-offs between No-SQL and Distributed SQL databases in cloud environments.

Adapted YCSB+T for benchmark across multiple database systems, enabling direct workload comparisons.

Achieved key milestones in benchmark, adapting YCSB+T for various databases, and initiating an open-source platform and educational resource development.

Conducted NLP and Machine Learning research to semantically match resumes to specific job descriptions or queries, aiding recruiters in finding the most suitable candidates.

Applied visualization output for embedding correlations, enhancing semantic analysis clarity.

Boosted recruitment efficiency by 25% with a tool that semantically matches resumes to job queries.

Delivered monthly progress presentations to the entire office and provided weekly internal updates.

Streamlined data workflow by optimizing the framework between Excel and Revit.

Cooperated with Mechanical, Electrical, and Plumbing teams for monthly value engineering initiatives.