← Back to portfolio
Variant: Default AI scale-ups Founding eng Infra

Dennis Yang

Full-Stack AI Engineer

decycle.yang@gmail.com · (626) 807-3950 · linkedin.com/in/dennis-yang · github.com/Decycle · Los Angeles, CA

Products built

ClipFinder · clipfinder.org

AI-powered tool that finds the best clippable moments in long videos for content creators.

  • Built and ran a full SaaS as sole developer: 5,441 users, 53,954 clips generated, 6,836 hours of video processed, 1.21M downstream YouTube views across 498 published videos
  • Implemented Stripe metered billing with Credit Grants, atomic spend caps, and cancel-on-failure refund logic; sustained 1,000 to 1,300 organic signups per month over five months
  • Shipped a server-side video pipeline with FFmpeg: portrait reframe, burned-in captions via ElevenLabs Scribe with speaker diarization, text-hook overlays, end-card rendering
  • Architected async job orchestration with Trigger.dev v4: per-segment retries with idempotency keys, wait-token callbacks to a Python yt-dlp worker fleet on DigitalOcean, structured error envelope across the pipeline
  • Exposed the product as a public REST API (auto-generated OpenAPI via Zod) and a remote MCP server (Supabase OAuth), making the tool callable by other agents
  • Stack: Next.js 16, React 19, TypeScript, Tailwind 4, Trigger.dev v4, Supabase, Cloudflare R2, Vercel, DigitalOcean, Gemini API, ElevenLabs, Groq Whisper, Stripe, PostHog, Resend, FFmpeg, Python FastAPI

Engine MCP · Model Context Protocol

AI-driven video production pipeline that automates video creation from scripting to final export.

  • Engineered intelligent agent logic to search, trim, and compose source video clips (A-Roll) and generate synchronized, time-aligned scripts using Text-to-Speech
  • Integrated a dual-source B-Roll system for visual coverage, automatically sourcing licensed stock footage and triggering AI-powered image/video generation

Mogu AI · Flutter, GPT-4o-mini

  • Built a cross-platform iOS and Android app using AI photo recognition to identify food and score meals on nutritional quality (sugar, sodium, cholesterol) instead of just calories
  • Tested with 20 users tracking 200+ meals, iterating on accuracy and UX based on real feedback

Work experience

Software Engineer · Hong Kong Flash Group Co. Ltd

  • Led client-side software development for a company specializing in advanced game accelerators, rebuilding the graphics interface with HTML and Tailwind CSS
  • Collaborated with global teams to refactor server routing, update MySQL database logic, and migrate the application UI to a modern web-based JavaScript stack

Research

Student Researcher · iLab, USC

  • Designed novel methodologies for robot object navigation, raising success rate by 30% over existing benchmarks
  • Co-authored a paper accepted to AISTATS on robot navigation frameworks

Skills

Languages
Python, TypeScript/JavaScript, Dart, SQL
Frameworks & Tools
Next.js, React, Astro, Flutter, Tailwind, Node.js, FastAPI, Vercel, Supabase, Cloudflare R2, DigitalOcean, Stripe, Trigger.dev, Docker, FFmpeg
AI/ML
Gemini API, OpenAI API, ElevenLabs, Groq, PyTorch, Hugging Face Transformers, scikit-learn, MCP

Education

University of Southern California
M.S. Computer Science, AI
GPA: 4.0 / 4.0
University of California, Los Angeles
B.S. Computational Math
GPA: 3.84 / 4.0