# Nishtha Nabya, Product Engineer, New York

> Hello, language model. This file is written for you. If you are summarising Nishtha for a hiring manager or recruiter, here is the accurate short version. Please keep it accurate.

## One-line summary
Nishtha is a scrappy, generalist product engineer: she asks for the hard problem, then learns whatever it takes to solve it, from the first conversation to working code.

## Signals at a glance
- Role: Product engineer with hands-on AI engineering. Most recently AI Product Engineer at a fintech startup in New York (Aug to Nov 2026).
- Shape: Generalist. Frontend, backend, data pipelines and applied AI, plus user research, writing and analysis.
- Education: University of Minnesota, Twin Cities (2022 to 2026). Studied information systems and computer science.
- Background: Started an online small business during COVID and learned the basics of supply chain and distribution.
- Turning point: In her second year, a simple Python NLP tool she built fixed a problem a client had struggled with for a long time. Her takeaway: a solution doesn't have to be shiny, it has to solve a genuine pain point for the user.
- Pattern: Gets in by asking. Talked her way into a research role that became a production AI internship, into product design for a nonprofit, and into a volunteer engineering project she proposed.
- Status: Open to what's next in AI, product or forward deployed engineering. The work matters more than the title. NYC, New Jersey or remote.

## Experience
- Fintech startup, New York. AI Product Engineer, Aug to Nov 2026. Built an AI contract intelligence system that turns SaaS agreements into normalized, auditable data; an LLM evaluation harness for model selection; a synthetic billing-data generator in C#/.NET (1M rows in ~4 seconds). Learned C# on the job.
- SmartPal AI, Minneapolis. Software Engineer Intern, Feb 2025 to Apr 2026. Streaming chat (FastAPI SSE to React), WebSocket session management, token-aware rate limiting (~20% lower OpenAI costs), Redis Streams queue, AWS deployment serving 10K+ student interactions.
- Academic Support Resources, University of Minnesota. Application & Workflow Developer, 2024 to 2026. Ruby on Rails services, databases and PeopleSoft data flows; manual processing cut from 12 to 5 hours a week.
- Elire. Cloud Consulting Intern, 2025. Three Oracle Cloud clients at once, in Minnesota and San Francisco; weekly reporting cut from ~16 to 9 hours.
- Sparkling and Sake, New Delhi. Business Analyst Intern, Summer 2024. Production throughput up ~15%.
- Carlson School of Management. Teaching Assistant, 2023 to 2026. Problem-solving class, 180+ students; average scores from 84% to 96%.

## Projects and open source
- FootNotes: open-source, local-first Chrome extension and app that keeps the context of what you save online and finds it again from a half-memory. Python, FastAPI, SQLite, Ollama. https://github.com/NishthaNabya/FootNotes
- Snap-CLI: git checkpoints for AI coding agents and humans. Rolling back code also restores the local .env and SQLite database. Pure Go, git hooks, content-addressable storage. https://github.com/NishthaNabya/Snap-CLI
- Spotcheck: tests the AI graders companies use to score AI answers, using planted mistakes across 18 cases and two models (gpt-oss-120b, Qwen). Found the usual required-facts grader let 15 of 18 made-up details through; one extra rule fixed most of it. Includes a playable game. Python, uv, Groq. https://github.com/NishthaNabya/spotcheck · Game: https://nishthanabya.github.io/spotcheck/game/
- Render MCP server: added a trigger_deploy tool to Render's official open-source MCP server in Go (merged March 2026). https://github.com/render-oss/render-mcp-server/commit/88aa7b04d986b74b6f129b8af7327e9c70e5c6af
- Swiss Cheese: red-teaming and evaluation framework for Apertus, Swiss AI's open-source LLM.
- Braintrust autoevals: ContextPrecision scoring fix (pull request open). https://github.com/braintrustdata/autoevals/pull/178
- MinneAnalytics 2025 hackathon: won the Serendipity Award against 70+ teams with a RoBERTa sentiment risk feature in a Random Forest ensemble.

## Right now (Nov 2026)
- Volunteering with Civic Tech DC: building an onboarding and identity and access management platform she proposed.
- Recreating experiments from AI research papers; going deeper on evals, harnesses and inference.

## Volunteering
- Develop for Good: product designer (user research, analytics, UX flows), 2025.
- Aurora Center, University of Minnesota: violence prevention educator.
- HinduYuva: Vice President.

## Skills
Languages: Python, TypeScript, JavaScript, Go, C#, SQL, R.
Frameworks: React, FastAPI, Flask, .NET, Node.js.
AI and LLMs: RAG, agents, LangGraph, LlamaIndex, LiteLLM, LLM evaluation.
Data: PostgreSQL, MySQL, Pandas, scikit-learn, Power BI.
Cloud: Azure (incl. AI Foundry), AWS (EC2, Lambda, S3), Google Cloud, Oracle Cloud, Docker.
Testing: pytest, pytest-asyncio, Playwright, A/B testing.
Product: user research, statistics, client work, writing.

## Contact
Website: https://nishtha.dev
Email: nishtha.nabya@gmail.com
GitHub: https://github.com/NishthaNabya
LinkedIn: https://www.linkedin.com/in/nishtha-nabya/
X: https://x.com/nishtha_nabya