Open to work · Available immediately →

Dennis Yang.

I build AI products end-to-end. The last five months I shipped ClipFinder solo, a SaaS for finding clippable moments in long videos. 53,954 clips generated, 1.21M downstream views, 5,441 users.

Looking for: founding engineer, product engineer, or full-stack AI roles. Based in LA, open to remote or relocating to SF.

M.S. Computer Science (AI), USC, May 2026 B.S. Computational Math, UCLA, 2024
Get in touch See the numbers

ClipFinder by the numbers

as of May 25, 2026

Pulled live from the production database. Solo build, five months, purely organic growth.

53,954
clips generated
6,836
hours of video processed
1.21M
downstream views
5,441
users
Scale
  • 6,720 jobs completed
  • 6,796 server-side video exports rendered
  • ~4.2% clip-level failure rate
  • 627 YouTube publishes via the tool
Growth
  • ~1,000 to 1,300 signups/month, five months running
  • 5.2% free-to-paying conversion (of active users)
  • $1,514 lifetime revenue, ~$496 in last 30 days
  • 1.21M downstream views, 498 videos tracked

Work

ClipFinder

Dec 2025 – Present

AI-powered video clipping for creators. Solo build, end-to-end.

TL;DR

I owned every layer: Next.js + React frontend, Trigger.dev task orchestration, a Python yt-dlp worker fleet on DigitalOcean, an FFmpeg server-side export pipeline, Stripe metered billing with credit grants and atomic spend caps, a public REST API, a remote MCP server, and email drips on Resend. No co-founder, no engineering hires.

What it does

Paste a YouTube, Twitch VOD, or Google Drive video. Gemini analyzes it and surfaces the moments worth clipping. You preview, trim, reframe to portrait, burn in captions, add a text hook, then publish to YouTube or TikTok directly. There's also a free AI clip-grader, a public REST API, an MCP server, and a Talk-to-Video chat over the transcript.

What I built

  • Async job orchestration via Trigger.dev v4. Per-segment retries with idempotency keys, wait-token callbacks to the Python worker, structured error envelope across the whole pipeline.
  • Server-side FFmpeg export. Face-aware portrait reframe, burned-in captions (ElevenLabs Scribe with speaker diarization), text hook overlay, end card. All rendered server-side, not in the browser.
  • Stripe metered billing with Credit Grants. Cancel-on-failure refund model. Atomic spend caps to prevent any single user from blowing past their cap. New users get a $2 grant on signup, no card required.
  • Multi-source ingestion. YouTube, Twitch VODs, Google Drive, direct upload. Pre-flight probes that throw structured errors before any meter event fires, so failed submits cost the user nothing.
  • Public REST API and remote MCP server (Supabase OAuth). The product is consumable by other agents, not just humans. Zod schemas generate the OpenAPI spec automatically.
  • 121 SQL migrations, RLS on every table, three test tiers (unit, integration, expensive), husky pre-push hooks. Built like a production system from week one.

What happened

From Dec 2025 to May 2026, ClipFinder processed 6,836 hours of video and generated 53,954 clips for 5,441 users. Users went on to post 498 of those clips to YouTube, where they collected 1.21M views and 24,383 likes. Top single video: 69,427 views. Sustained roughly 1,000 to 1,300 signups per month for five straight months on organic alone.

Revenue tells a different story: $1,514 lifetime, ~$496 in the last 30 days. That gap, scale up and revenue flat, is the lesson.

“We'd given up on clipping after seeing how long it took to select, caption, and export ourselves. Since switching, we've reached over 20,000 new YouTube viewers in a month, and publishing clips is actually fun.”
Swtorista
Gaming YouTuber · 181K subscribers

What I learned

The honest read: most short-form creators don't earn enough from shorts to be reliable customers. That's a structural ceiling on willingness-to-pay, not a product problem. Most of my customers used the tool trying to "be famous with shorts." Most don't make it. Their churn is my churn.

I'd rather work on a bigger problem at a place with more leverage than keep optimizing for a constrained segment. ClipFinder keeps running, the infrastructure takes care of itself, and I'm looking for what's next.

Earlier work

Stack

Tools I've shipped production code with. Not a wishlist.

Product
Next.js, React, TypeScript, Tailwind, Astro
Infra
Vercel, Cloudflare R2, DigitalOcean, Trigger.dev, Docker
Data
Postgres, Supabase, Row-Level Security, Upstash Redis
AI
Gemini API, ElevenLabs Scribe, Groq Whisper, OpenAI API, PyTorch
Payments + Email
Stripe metered billing, Stripe Credit Grants, Resend
Observability
PostHog, Microsoft Clarity, Vercel Analytics
Languages
TypeScript, Python, SQL, Dart

Path

Dennis Yang

Two years ago I was a math undergrad at UCLA. One year ago I was starting an AI MS at USC. Last month I graduated. Somewhere in between I shipped a SaaS.

Before all that: a Hong Kong summer rebuilding a game-accelerator UI in HTML and Tailwind, two years researching robot navigation at USC's iLab (paper accepted to AISTATS), and a string of side projects that taught me how the whole stack fits together.

I work across the stack on purpose. Frontend, backend, infra, AI pipelines, billing, growth instrumentation. ClipFinder was the first time I had to do all of it at once, for the same product, for real users who'd actually email me when something broke. It changed how I think about engineering.

Now

May 2026. Just graduated from USC. I'm reading job posts, sketching ideas for what's next, and using most of my time to talk to people who are building things I'd want to be part of.

If you're hiring for a founding engineer, product engineer, or full-stack AI role, I'd like to hear about it.

Contact