I build risk-aware forecasting and streaming algorithms.
PhD researcher (NJIT) working on time/space-efficient methods for streaming data, with industry experience in ML, risk, and analytics.
About
I work at the intersection of algorithms, machine learning, and systems — especially where correctness and performance matter. I'm currently building a product around risk-aware forecasting for equities that makes tail risk visible and actionable for individual traders.
Interests: streaming & approximation algorithms, quantiles/sketches, risk metrics (CVaR), calibration, and real-time analytics.
Now building
Product-heavy summer project focused on individual equity traders.
Trading Strategy Explorer
Interactive explorer covering six strategy families — trend/momentum, mean reversion, factor, HFT microstructure, options/volatility, and event-driven — with math, live charts, and sample runs on real market data.
Stock prices and VIX pulled from Yahoo Finance and updated automatically each weekday after market close.
Quantile Sketch Tracer — Interactive Demo
Interactive visualization of streaming quantile sketch algorithms (e.g., GK, t-digest, KLL). Step through inserts, watch the sketch state evolve, and compare error/space trade-offs in real time.
Risk-aware forecasting for equities
A decision-support tool that forecasts the distribution of returns (not just a point estimate), surfaces tail risk (CVaR), and flags when risk expands or regimes shift.
- Quantile fan chart + CVaR + drawdown probability
- Risk change alerts with plain-English explanations
- Stress-period playback (e.g., 2020, 2022) to build trust
If you’re a trader or builder who wants to test this, email me — I’m recruiting early users.
What I’m looking for
Feedback from active equity traders and engineers who care about reliability, calibration, and real-world constraints.
- What risk views do you actually use (or wish existed)?
- Which horizons matter (5d / 10d / 20d)?
- What “alerts” would change your behavior?
Portfolio
Selected projects and papers.
Computing Estimators of a Quantile and CVaR
Benchmarks sorting versus selection algorithms for computing two core risk measures — quantile and CVaR — on pre-generated simulation data. Finds that selection consistently beats sorting, with the fastest strategy depending on dataset characteristics.
Sentiment Analysis on Rotten Tomatoes Movie Reviews
NLP research project exploring sentiment classification for movie reviews, including baselines and feature-driven improvements.
Covid-19 Sentiment & Cases vs. Amazon / Walmart Preference
Used big-data tooling + NLP/ML to study how covid case trends and public Twitter sentiment relate to stock/market preference signals.
If you have a dedicated write-up or PDF for this, swap the “Details” link to it.
Selected experience
A few highlights (full details on LinkedIn).
Time/space-efficient algorithms for streaming data; benchmarking and systems prototypes in C++/Python.
Predictive modeling for commercial outcomes using large-scale entertainment platform metadata.
Built risk/exposure models, explored trading strategies, and developed ML-based signals.
Contact
Best way to reach me is email.