MGNY Consulting Corp.

About the project
We’re working with a US-based (New York) proptech/govtech startup building the system of record for NYC affordable-housing compliance — a platform that takes a housing project from approval, through the lottery, to staying compliant year after year. Their AI review pipeline is already in production, processing real applications for real buildings. Lots of messy real-world documents, deterministic regulatory logic, and an AI layer that does the volume while experts handle exceptions.

It’s a long-term, stable cooperation with a small, high-ownership, high-collaboration team.

Cooperation

• GIG or B2B contract
• Full-time, long-term

• Remote, US-market focus

   

What you’ll do

• Own features end-to-end: spec, data model, migrations, backend services, pipelines, and UI
• Build LLM-powered document intelligence (extraction, classification, gap detection) over messy real documents — scans, spreadsheets, handwriting
• Design and run eval loops that make AI shippable: golden datasets from expert feedback, regression guards in CI
• Draw the line between AI and deterministic code — LLMs classify/extract/suggest; anything filed with a regulator runs through pure, unit-tested policy logic

• Ship Python services (FastAPI, Postgres, event-driven) and React/TypeScript frontends

   

Requirements

• 5+ years building and operating production software (senior product engineer)
• Strong Python (FastAPI, Pydantic, SQLAlchemy, Postgres)
• Real, hands-on LLM/GenAI experience: prompt pipelines with structured outputs, evals, OCR/document processing, agents — features that survived real users
• Frontend with React + TypeScript
• AI-native workflow: you use AI/coding agents to write most of your code and know how to verify their output
• Comfortable with ambiguity and defining fuzzy problems yourself
• Rigorous engineering: TDD, audit trails, verification

• English B2+ (comfortable communicating with the team)

   

Nice to have

• Proptech, fintech, govtech, or legaltech (regulated, document-heavy domains)
• Event-driven architecture (outbox pattern, streams, idempotent consumers)
• Experience migrating a business off a legacy system while it stayed live (e.g. Salesforce)

• Built eval infrastructure or human-in-the-loop review systems

   

Interview process

1. Hiring manager screen (background, ownership, how you work with AI)
2. Coding screen — a realistic task from the domain (bring your AI tools)
3. AI system design — a loosely defined real problem
4. CEO conversation