San Jose · CA
Portfolio · v.2026
Kevin Cui
Full-stack software engineer.
Building thoughtful products with intelligence quietly inside.
Full-stack software engineer.
Building thoughtful products with intelligence quietly inside.
I’m a Dec 2025 Data Science grad from San Jose State University, building full-stack AI products where customers have a need and I have an answer.
I’m now working on JanusLabs (an AI marketing tool that started as a GTM Hackathon finalist and now sells direct to consumers), a browser-based 3D classroom with persona tutors, and a plain-language medical assistant. Each one pushed me further into system design and AI engineering that I wish I’d learned sooner. I’m open to full-time conversations, or opportunities.
An AI marketing platform I'm building and selling to consumers. Started as a GTM Hackathon finalist; rebuilt post-hack with an Apify-cached scraping pipeline, multi-provider AI rendering, and Stripe billing. Scrapes TikTok, Reels, and Shorts trends, turns them into on-brand scripts, generates AI video. Still small but actively shipping.
A browser-based social AI classroom. Learners move through a shared 3D space, talk to persona-based tutors that can be shared or duplicated, and join the same live room with a code. Agent flows use reinforcement learning with Redis-backed memory.
Plain-language medical guidance with voice transcription and optional TTS, wiring multipart and JSON flows through Next.js App Routes. MiniMax powers selectable prompts for the user to choose from.
Sentiment analysis for bubble tea shops, fine-tuned on 6M+ Yelp reviews. PostgreSQL model with sentiment scores, labels, and indexes for sub-200ms loads. Frontend on Vercel, backend on Railway.
Dash + Flask on GCP, built in a 2-person team. Python pipelines scraped and processed 100+ competitive battle files across 10+ years. A Random Forest model recommends movesets with ~70% accuracy.
Cloud-hosted finance tracker (Flask + Dash). User auth and persistent data in Google Cloud Storage, real-time Plotly dashboards, and exception handling for stable runtime.
Always up for a conversation about thoughtful software, AI products, data systems, or any project where the model and the interface have to think as one piece.