Life
Born in New Zealand, raised in Hyderabad
Fifteen-plus countries so far, chasing culture, people, and perspective.
Auckland → Hyderabad → Minneapolis → New York
New York local
Prashant Pilla
Software Engineer — AI systems for finance
I like building systems where new technology has to survive contact with real money and real people.
01 — Highlights
Things I have built, and a few I have lived.
Builds, research, ventures, and life. Press f to cycle the filter.
Showing 12 highlights
- Research
Forecast the S&P 500 with LSTMs at 96.41% accuracy
Compared ARIMA and LSTM networks on a decade of index data. The feature-free LSTM won with an MAE of 175.9.
arXiv · Jan 2025
- Build
Colosseum Breakout Hackathon finalist
Tabi: Solana payments, group bill splitting, and an AI advisor on your wallet, shipped as a working MVP in under 48 hours.
Solana · May 2025
- Build
An insurance-claims agent where the LLM cannot skip a step
A deterministic controller owns the procedure; the model only extracts and phrases. Twelve test suites, CI, Docker.
Python · CI · Docker
- Venture
First investment at ADIN
Two months into scouting for the Autonomous Deal Investment Network, the AI-powered venture engine backed its first company.
Nov 2025
- Build
A campus drone delivery sim in C++ and Three.js
Priority queues, a shared wind field, leader/helper battery handoff, and A* routing over the UMN map. Dockerized.
C++ · Three.js · Docker
- Life
Certified yoga instructor at seventeen
Sivananda ashram in the Himalayas, where the lineage began. One of the youngest internationally certified.
Jul 2022
- Build
Palantir Foundry certified, twice
Foundry Aware Professional and Foundry & AIP Builder Foundations: ontology-driven apps, enterprise data modeling, operational workflows.
Jun 2026
- Life
Six languages, counting English
Telugu, Hindi, Urdu, Spanish, and French. The preloader says hello in all of them.
- Research
Dean's List, three years running
University of Minnesota, 2022 through 2024, alongside the MLK and MCAE scholarships.
UMN · 2022–2024
- Venture
Outreach Coordinator, UMN Blockchain Club
Ran events and workshops and built industry partnerships to bring blockchain to the wider campus.
2024–2025
- Life
MLK Scholar
University of Minnesota. Also a CSE and CLA Student Ambassador on 10+ student panels.
UMN
02 — Selected Work
Things I have shipped.
Case studies with the problem, what I built, and what happened.
- 01TabiSolana payments, group bill splitting, and an AI advisor on your wallet. Colosseum Breakout 2025 finalist.Next.js 15TypeScriptExpressSolana Web3.js2025
- 02open-verseLLM trading agents compete on paper markets. An orchestrator ticks the market, calls the models, places paper orders, and a dashboard shows who is winning.TypeScriptNode.jsOpenAI SDKAlpaca Paper Trading2025
- 03Gopher Drone Sim3D campus drone delivery simulation. A C++ engine with priority queues, weather, and leader/helper handoff talks to a Three.js frontend over WebSockets.C++libwebsocketsMakeTypeScript2025
- 04CiteworkGrounded investment-committee memos. The model extracts, Python does the math, and a human approves before a sentence of prose exists.Python 3.12Anthropic APIStreamlitDecimal arithmetic2026
- 05Insurance Claims SOP AgentA conversational insurance-claims agent where a deterministic controller enforces the procedure and the LLM only extracts and phrases.Python 3.12FastAPIOpenAI / Anthropic APIsDocker2026
- 06ADIN Marketing SiteNext.js marketing site for ADIN, an AI-first venture DAO, with dedicated flows for investors, members, and founders.Next.js 15React 19TypeScriptTailwind CSS 42025
03 — About
Born in Auckland. Raised in Hyderabad. Building in New York.
I build AI systems that have to survive contact with real money. Recent work includes production AI for institutional investors at Tribute Labs in New York, scouting early-stage companies for ADIN, and a run of open-source systems: an insurance-claims agent harness where a deterministic controller enforces procedure, Citework for grounded investment memos, and open-verse, an arena for LLM trading agents.
I was born in New Zealand and grew up in Hyderabad. Somewhere between the two I picked up a habit of paying attention to how people from different places think, which is probably why I ended up caring more about who a system is for than what it is made of.
At seventeen I became a certified yoga instructor at a Sivananda ashram in the Himalayas. It taught me that discipline is mostly about showing up quietly and often, and that is still how I approach code, research, and the occasional 36-hour hackathon.
I studied computer science at the University of Minnesota, where I published a deep learning paper on forecasting the S&P 500 and spent too many evenings at the Blockchain Club. I am happiest when a problem sits between engineering, product, and the people who have to trust the result.
- Focus
- AI systems for finance: agents, grounded analysis, trading infrastructure
- Education
- B.S. Computer Science, University of Minnesota
- Certifications
- Palantir Foundry Aware Professional; Foundry & AIP Builder Foundations; Yoga Instructor Level I (Sivananda)
- Cities
- Auckland, Hyderabad, Minneapolis, New York
- Languages
- English, Telugu, Hindi, Urdu, Spanish, French
Research · arXiv · Jan 29, 2025
Forecasting S&P 500 Using LSTM Models
Prashant Pilla, Raji Mekonen
Compares ARIMA and LSTM networks for forecasting the S&P 500 on daily data from October 2013 to September 2024 with Bloomberg technical and macro features. The LSTM without additional features performed best, and the paper argues that sequence models handle non-linear market dependencies that linear baselines miss.
- LSTM accuracy (no features)
- 96.41%
- LSTM MAE / RMSE (no features)
- 175.9 / 207.34
- LSTM accuracy (with features)
- 92.46%
- LSTM MAE / RMSE (with features)
- 369.32 / 412.84
- ARIMA accuracy
- 89.8%
- ARIMA MAE / RMSE
- 462.1 / 614
04 — Ventures
Engineer, and builder.
Scouting, angel work, and the occasional company. If you are a founder or want to build something together, my inbox is open.