I'm a computational scientist building and deploying full-stack production-grade machine learning systems. I hold a PhD from the University of Texas at Austin, where I developed a strong foundation in numerical methods, stochastic modeling, and scientific computing. Today I apply that rigor to quantitative finance, optimization, anomaly detection, and NLP systems that need to work reliably in production.
I've shipped automated trading systems, RAG-based analytics platforms, structured-data Q&A agents, and daily performance dashboards. I care about the full lifecycle: feature engineering, model development, experiment design, deployment, evaluation, and monitoring.
I built production ML systems and trading infrastructure for quantitative strategies. That included volatility arbitrage intraday options strategies on AWS Lambda, constrained nonlinear optimization for stochastic volatility model calibration, and NSGA-II/Pareto-front portfolio construction for sector ETF allocation with regime-adaptive rebalancing.
I also worked deeply on model development and experimentation: optimizing options-chain backtesting pipelines with Pandas, deploying Hidden Markov Models with Bayesian Information Criterion for regime detection, and running A/B tests comparing Shapley-based and model-based feature attribution methods for feature selection. On the sequence-modeling side, I deployed Chronos embeddings, LoRA tuning workflows, and AutoGluon-backed trading systems on Modal and AWS Lambda.
Beyond trading systems, I built multiple NLP analytics products. For unstructured data, I delivered a RAG-powered PDF Q&A application with Streamlit, Groq-hosted Llama models, and ChromaDB/Qdrant vector backends, including ensemble retrieval, deduplication, reranking, and dual-LLM answer refinement. For structured data, I built a Groq-based Q&A agent with a SQL-backed connector layer, LLM-based planning and evaluation, and domain-aware query correction.
I rounded this out with a FastAPI and Streamlit explainer microservice for US Treasury press releases using Groq and FAISS, and production safeguards including semantic injection detection, Presidio-based PII redaction, Guardrails AI output enforcement, LangSmith tracing, and offline evaluation. I also built React performance dashboards backed by DynamoDB for day-to-day monitoring.
I joined during the early integration phase of a robotic surgical navigation platform that combined medical imaging with robotic actuation. My work focused on implementing TCP/UDP communication protocols for robotic arm control, contributing to end-to-end packaging with Poetry, and supporting deployment through Azure tooling. The team delivered a successful live system demonstration that helped support the company's Series A fundraising effort.
I adapted physics-informed modeling originally developed for fusion experiment data to financial time series. The resulting production pipeline combined sparse regression for PDE construction, automated signal generation, and automated execution. I also implemented a CAGR-maximizing Bayesian TPE optimizer for swing-trading thresholds and deployed the system on AWS Lambda with CloudFormation, Docker, GitHub Actions, and EventBridge scheduling.
ML & Optimization
Model development, feature engineering, and optimization across classical ML, probabilistic models, and evolutionary search.
Cloud & MLOps
Serverless-first production systems with CI/CD, infrastructure-as-code, containers, and model deployment workflows.
Programming
Python-first engineering with scientific computing, fast data processing, and frontend work for internal analytics tools.
APIs & Web
Production APIs and app backends connecting market data, broker integrations, and interactive analytics interfaces.
NLP & LLMs
RAG systems and LLM tooling with guardrails, observability, vector retrieval, and evaluation for production use.
Beyond the code
