Engineering that understands the business.
Forward Deployed · Applied AI Engineer

Mat Niznik
Forward Deployed · Applied AI Engineer · Long Beach, CA
M.Sc. Computer Science, Westcliff University Irvine · MBA, Bethesda University of California · M.Sc. E-Business, SGH Warsaw School of Economics
Applied AI engineer in Long Beach, CA. I put AI into real workflows and own the result: the model takes the first pass at scale, tests sit behind it, and a person makes the calls that carry real stakes.
Ten years shipping web and mobile in TypeScript, React, React Native, Node.js, Python, Flask, Django, PostgreSQL and AWS. I also teach programming at Bethesda University of California. Open to Forward Deployed and Applied AI engineering roles.
- I'm Mat, a forward deployed / applied AI engineer in Long Beach. Ten years across web, mobile and games. M.Sc. Computer Science, MBA, and I teach at Bethesda University.
- I put AI into the work people already do. The product, the AI inside it, and the checks around it.
- It's already in production. At MintCollect I rebuilt an app with 85K+ accounts and ran its production. My Claude agent there passed 210 of 210 eval answers.
- I start with the people doing the work. We map the workflow together. AI goes where it adds judgment, and a person approves anything that leaves.
- Open to Forward Deployed and Applied AI roles. Or bring the workflow that eats your week to a 15-minute call, and we'll see if AI belongs in it.
What I build with
- Claude
- Amazon Bedrock
- Claude Code
- Codex
- MCP
- LangGraph
- RAG
- Human-in-the-loop
- Evals
- Cursor
- TypeScript
- JavaScript
- React
- Redux
- Vite
- Tailwind CSS
- HTML5
- CSS
- React Native
- Expo
- RevenueCat
- Python
- Node.js
- Express
- Flask
- Django
- FastAPI
- SQLAlchemy
- Gunicorn
- Streamlit
- PHP
- Laravel
- Java
- Spring Boot
- C++
- PostgreSQL
- MySQL
- MongoDB
- Elasticsearch
- Redis
- Firebase
- Kafka
- Snowflake
- dbt
- Power BI
- AWS
- Azure
- Cloudflare Workers
- Terraform
- Docker
- Kubernetes
- GitHub Actions
- Git
- Nginx
- Sentry
- pytest
- WebdriverIO
- Microsoft Clarity
- Google Analytics
- Stripe
- Shopify
- Salesforce
- Gmail API
- Figma
- Jira
- Miro
- Postman
- VS Code
- Unity
- C#
- Photon Fusion
Where I used them: ledger-desk · recall-desk · rag.irs · BUC Populi · Niznik Studio · MintCollect · MatOS · SentiKits · Zippia · Taycor · Streetvers
Tap a circle, or the dot where they meet
The overlap
Code. Business. Customers. One engineer.
- Six department chairs trained to use AI, starting with the fear, not the feature
- A testing framework that made product iteration 42% faster at Zippia
What I can build for you
Twelve things I have built, grouped by where they help. Each one has proof from my own work.
AI drafts, a person decides
Grading, reviews and compliance need judgment and a name on them.
Claude drafts against your own rubric, code does the fetching and the rules, and nothing goes out until a person approves it. Then the system reads it back.
My weekly grading pipeline for 5 courses: 43 scripts, every grade approved, then read back from Populi.
Selected Work
Products I've built and run - the AI, the app around it, and the checks that keep it honest

Claude reads each supplier bill, code checks it against the order, the delivery and the bank details, and every approval is a signature the books verify
ledger-desk.mniznik.com ↗
A support agent that answers from NHTSA's public data with seven tools, and six code rules check every answer before a customer sees it
recall-desk.mniznik.com ↗
Answers tax questions from 11 IRS publications and checks every quote word for word, or says "not found"
rag.irs.mniznik.com ↗


Rebuilt the iOS and Android app in React Native and became its only app engineer: 22 versions in seven months, for an app with 85K+ accounts
Read the case study →
AI agents that run my working life, with a human gate on everything that leaves
Read the case study →Claude drafts each grade against the rubric; nothing posts until a person approves it
Read the case study →
A multiplayer card game I built in Unity, and the API behind it
Read the case study →


A wellness app with an in-app shop, 30 Cloud Functions and server-checked Apple receipts
Read the case study →Things I did before I knew better

The walk-in pitch
Fun fact - at 20 I walked into the Sequoia Capital office in Menlo Park with no meeting and one brilliant idea.
ALLSPOT. Pokemon Go meets GTA.
I asked if I could talk to someone about it.
They told me to book a meeting first.
Fair.
So I buckled up and got to work. One day I'll book that meeting.

How it starts · A 15-minute call