Saumik Dana | ML Engineer & Quantitative Researcher

About

I build ML systems that actually make money. Three years deep in quantitative finance, specializing in turning messy market data into systematic strategies. From Chronos embeddings to regime detection, I handle the full stack—modeling, optimization, deployment, monitoring. PhD background in computational mechanics gives me an edge in algorithmic thinking and numerical methods.

What I've Built

Feb 2024 – Dec 2024

Quantitative Researcher

Asset Management Firm – Stamford, CT

Built serverless trading infrastructure on AWS (Lambda, EventBridge, S3, DynamoDB) running automated options strategies. Deployed foundation models for time-series feature extraction, HMM for regime detection, and multi-objective optimization for portfolio construction.

Backtested performance: 1.5+ Sharpe, 10% max drawdown, 10% annualized alpha

Created LLM-powered analytics using RAG to parse treasury policy releases and assess market impact. Built real-time React dashboards tracking volatility surfaces and P&L with sub-second latency.

Optimized data pipelines with vectorization and JIT compilation, achieving 100-1000x speedups on options chain processing. Used SHAP for feature attribution, Bayesian optimization for hyperparameters, and genetic algorithms for signal discovery.

Aug 2023 – Nov 2023

Computational Engineer

VISIE Inc. – Austin, TX

Early-stage surgical robotics startup. Implemented real-time control protocols (TCP/UDP) for robotic arm motion and built Azure deployment pipelines. Contributed to successful Series A demo.

Aug 2022 – Mar 2023

Computational Lead

Sophelio – Austin, TX

Applied physics-informed ML (originally designed for fusion plasma experiments) to financial time series. Built production pipeline using sparse regression for PDE discovery and Bayesian optimization for signal thresholds. Deployed on AWS Lambda with automated execution.

Tech Stack

ML & Optimization

PyTorch, XGBoost, SHAP
Bayesian optimization (Optuna, TPE)
NSGA-II, genetic algorithms
HMM, RL (Conservative Q-Learning)

LLMs & NLP

LangChain, RAG pipelines
FAISS, HuggingFace embeddings
Chain-of-thought prompting

Cloud & Infrastructure

AWS (Lambda, S3, DynamoDB, EventBridge)
Docker, GitHub Actions
CloudFormation, serverless architecture

Programming

Python (Pandas, NumPy, JAX, Numba)
React, FastAPI
Git, Linux/Bash

Education

PhD in Engineering Mechanics

University of Texas at Austin

Developed multiscale numerical frameworks for coupled flow and geomechanics. Published in Journal of Computational Physics. Postdoc at USC.

MS in Mechanical Engineering

Indian Institute of Science, Bangalore

BS in Mechanical Engineering

University of Mumbai

Beyond the Code