I'm a computational engineer who designs, ships, and operates full-stack ML and LLM-based systems end-to-end. My PhD at the University of Texas at Austin grounded me in numerical methods, stochastic modeling, and scientific computing — a toolkit I now turn on quantitative finance, retrieval-augmented generation, and multimodal document intelligence.
In practice, that means production trading strategies, LLM-backed research and discovery agents, hybrid retrieval pipelines, and the cloud-native scaffolding around them. I care about the whole lifecycle — feature engineering, model selection, calibration, deployment, evaluation, and the boring infrastructure that keeps any of it from breaking at 3 a.m.
I built and ran production ML and trading infrastructure on a serverless-first stack — AWS Lambda, DynamoDB, S3, EventBridge, CloudFormation, and Modal — wired together with a GitHub CI/CD pipeline using Docker buildx and an ECR registry cache, plus React dashboards over the DynamoDB/S3 state and lifecycle cleanup of stale ECR images. On top of that foundation I shipped an anomaly-detection-backed volatility arbitrage strategy for intraday options, a LoRA-tuned Chronos and TabPFN driven interday options system, and a bi-objective optimizer backed thematic equity portfolio with regime-adaptive rebalancing.
On the LLM side, I evaluated Groq-backed agents with RSS/XML ingestion, guardrails, and structured trading-signal outputs; engineered FX and macro-news driven intraday FX signal evaluation agents; and benchmarked hand-rolled prompt architectures against LangGraph for FX trade position sizing. I also built a ticker alert pipeline with hybrid retrieval — SQLite FTS5, PostgreSQL tsvector, Whoosh, and FAISS — feeding RAG, along with a query-based signal mining Streamlit app over backtested options strategies and a multimodal embedding model wired to a Next.js PDF Q&A app for market research.
Behind all of that was a steady stream of experimentation: bi-objective optimization with different Pareto-front selection rules, A/B tests comparing SHAP versus native feature importance, Black–Litterman portfolio construction, fast jump-diffusion-aware Heston model calibration via constrained optimization, and OOS benchmarks pitting Google TimesFM against Amazon Chronos and LoRA-adapted against base Chronos.
I joined during the early integration phase of a surgical navigation platform combining imaging and robotic actuation. My work centered on implementing TCP/UDP communication protocols for robotic arm motion control, supporting end-to-end product packaging and deployment, and helping stage a live demonstration that anchored the company's successful Series A round.
I adapted physics-informed modeling originally developed for fusion experiment data to financial time series. The resulting production pipeline used sparse regression for PDE construction and signal generation, paired with a CAGR-maximizing Bayesian TPE optimizer driving a swing-trading system deployed on AWS Lambda.
ML & Optimization
Model development, feature engineering, and optimization across classical ML, probabilistic models, time-series foundation 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 & Dashboards
Production APIs and app backends connecting market data, broker integrations, feeds, and interactive analytics interfaces.
LLM & RAG
LLM tooling and hybrid retrieval systems combining lexical and vector search for production RAG.
Beyond the code
