Software & AI Engineer · New York

Samuel Jo

I build living products — software that ships, then gets used every day.

Full-stack and AI engineer. I take products from blank repo to production: type-safe interfaces, real-time game servers, civic-data pipelines over millions of records, and LLM features kept honest by the data behind them.

01 Selected work

Five products. All live, all in use.

01 Live

Crowdtells

AI · News · Prediction Markets

A newsroom driven by the money. It reads Polymarket and Kalshi as a signal of what matters, ranks stories by newsworthiness, then briefs each one from many outlets with LLM summaries kept strictly grounded in the market data.

ReactTypeScriptGroq LLMCloudflare
02 Live · Founder

RegWatch

SaaS · Civic Data · Full-Stack

Property intelligence for NY, NJ & CT. Instant compliance and risk reports over 10.9M+ parcels drawn from 273 government sources, with real-time monitoring, grounded AI chat, and a 12-language UI — all from one multi-brand codebase.

Next.js 15React 19SQLiteStripeGroq
03 Live

facecard

Live-Face Games · 3D · WebRTC

A browser live-face table-games platform — Poker, Mafia & Tonk — where each player’s real webcam face becomes a live 478-point video-textured mesh around the table. An authoritative server runs the games; faces and spatial voice stream peer-to-peer over a LiveKit SFU.

TypeScriptThree.jsMediaPipeLiveKitSocket.IO
04 Live · Private

miztips

PWA · Mobile-First · Real-Time

A tip-distribution PWA run every shift at a NYC restaurant — role-weighted pooling, shift tracking, branded PDF exports, and real-time sync with an offline fallback. Installable, touch-first, and trusted with people’s pay. It runs on real staff data, so it’s password-protected — the case study is the way in.

ReactFirebaseTailwindjsPDFPWA
05 Live

phos-analysis

Equity Research · Data Viz · Front-End

An interactive equity-research report on First Phosphate Corp. (CSE: PHOS), read straight from its SEDAR+ filings — a live share price, a bear-to-bull fair-value range, PEA economics, and ten sections from thesis to ESG. A research report that behaves like software.

JavaScriptData vizLive market dataVercel

02 The engineer

About

I’m a software and AI engineer in New York. I like building things people actually use, and taking them all the way to production — not just to a demo.

My work spans the whole stack: type-safe frontends, real-time game servers, large civic-data pipelines, and LLM features engineered to stay grounded in real data. I care about the parts that never show up in a screenshot — test coverage, offline behavior, accessibility, and interfaces that feel considered down to the last pixel.

Lately that has meant a prediction-market newsroom, a property-intelligence platform over millions of parcels, a real-time 3D poker table, and a tip calculator a restaurant runs on every night.

Based
New York · —:—
Focus
AI & full-stack engineering
Shipped
5 live products
Now
Building RegWatch

03 Capabilities

What I reach for.

AI Engineering where I focus

Production LLM systems that run unattended and stay correct at scale — grounded, evaluated, and shipped.

LLM orchestration Claude · GPT · Groq · Llama RAG architecture Prompt engineering Structured output & guardrails Few-shot · chain-of-thought Autonomous agents Agentic tool-use Browser automation · Playwright Grounded generation NLP — summarize · classify Evals & eval harnesses Ranking & MMR MediaPipe vision Streaming · JSON mode

Languages6

  • TypeScript / JavaScript
  • Python
  • C#
  • SQL
  • HTML & CSS
  • Bash

Frontend7

  • React / Next.js
  • Vite
  • Three.js
  • Tailwind
  • Pretext
  • PWAs
  • Accessibility

Backend & Infra9

  • Node / Express
  • Socket.IO · WebRTC
  • SQLite · Postgres
  • Firebase · Supabase
  • Stripe
  • Cloudflare · Vercel
  • Render · VPS / PM2
  • CI/CD · GH Actions
  • Playwright · Vitest

Data6

  • Data pipelines / ETL
  • FTS / search
  • Schema design
  • WAL tuning · indexes
  • i18n at scale
  • Cron / scheduled jobs

04 Play

Deal me in.

You vs. two bots that bet by live Monte-Carlo equity — the engine behind facecard.

100 chips Ante 5 Pot 0
Board
You

You vs. two opponents · equity & showdown run in your browser · see the real engine →

05 — Contact

Let’s build something.

Email is the fastest way to reach me — I read everything. If you’re building something interesting, I’d like to hear about it.

or press anywhere to jump around