Things I built end to end — mostly systems where the data had to be right and the latency had to be low.
2026
Insight for Impact
An AI-powered humanitarian command center built for Hacklytics 2026 (Databricks × United Nations). It turns fragmented UN humanitarian data into something you can actually act on: a 3D crisis globe, funding gap analytics, ML forecasts, and natural-language queries over live data.
Two-stage ML pipeline — Prophet for funding trends, XGBoost for needs prediction
Interactive 3D globe with Globe.gl / Three.js and Plotly analytics
Conversational queries over UN HNO/HRP data via Databricks Genie
Mismatch scoring to surface high-neglect-risk regions
Python
XGBoost
Prophet
Streamlit
Built with Nikil Kandala for Hacklytics 2026: Golden Byte at Georgia Tech.
A full-stack order entry system for hedge fund traders, built with a three-person team. Redis caching and low-latency matching over FastAPI and WebSockets get request/response under 50ms, behind a deliberately lightweight HTMX front end.
Sub-50ms request/response on local benchmarks
Redis sorted sets for price-time priority, atomic matching via Lua scripts
Real-time order book streamed over WebSockets
Risk limits, role-based access, multi-account support, and audit logging
Python
FastAPI
Redis
Demoed at Millennium HQ in New York — order submission, real-time matching, and risk controls.
Designed the EERD and built an optimized SQL database for a cruise line management system, with the constraints and stored procedures needed to keep the data honest under complex booking scenarios.
Optimized constraints and stored procedures
Enforced data integrity without sacrificing query performance
Schema designed to scale with operational complexity
A workplace safety app that reads live air quality from PurpleAir sensors and recommends protective equipment against safety thresholds, sourcing real products through the Amazon Product Advertising API.
Real-time air quality monitoring across sensor APIs
Automated PPE recommendations driven by safety thresholds
Dashboard for workplace safety metrics
Python
Javascript
Node.js
—
Air Quality Analysis
A statistical study of more than a million pollution data points, using correlation analysis, time-series modeling, and forecasting to find the trends underneath the noise.
Correlation analysis across a million-plus readings
Python
FastAPI
Redis
SQL
Javascript
Node.js
Numpy
Pandas
MatPlotlib