Siddharth Chilamkur

Siddharth Chilamkur

About Me

I'm a machine learning and optimization enthusiast with extensive experience in finance and aerospace applications. I thrive at the intersection of theory and practical systems — building intelligent pipelines, forecasting models, and scalable infrastructure with real-world impact.

Education

University of California, Berkeley – B.S. Electrical Engineering & Computer Science (2026)

B.A. Economics (2026)

Relevant coursework: Machine Learning, Artificial Intelligence, NLP, Agentic AI, Computer Vision, Optimization, Robotics, Database Systems, Financial Economics, Antitrust Economics, Econometrics, Industrial Organization

Professional Experience

National Aeronautics and Space Administration

NASA

AI/Engineering intern since June 2023, working on systems modeling and optimization for exploration technology as well as predictive analysis for flight trajectories, ATC autonomy, and climate modeling.

Optimization Research

Optimization

Working on optimization research with Prof. Hochbaum (IEOR @ UC Berkeley) in the areas of active learning, anti-clustering, and scenario reduction techniques with applications to real-world systems.

Selected Projects

Equity Strategy

Equity Strategy Backtest

Simulated long-short portfolios based on ML signals. Evaluated with Sharpe, Sortino, and Calmar ratios in a production-ready backtesting pipeline.

Earnings Sentiment

Earnings Sentiment & Return Forecasting

Used NLP and transformer embeddings to extract sentiment from earnings calls and predict near-term returns using regression models.

Financial Fraud Detection

Financial Fraud Detection

Developed a real-time ensemble ML pipeline with 92.3% recall and 0.983 AUC, designed for enterprise-scale risk detection and compliance.

Tire Degradation

F1 Tire Degradation Modeling

Regression model with SHAP-based interpretability to predict tire performance drop-off using telemetry data from Formula 1 races.

MLB Prediction

MLB Pitch Outcome Prediction

Two-tier gradient boosting classifier trained on Statcast data to predict pitch outcomes with interpretable feature importance.

NeRF

Neural Radiance Fields (NeRF)

Trained volumetric rendering models to synthesize new 3D views from 2D images using positional encoding and MLPs.

Diffusion

Diffusion Modeling

Implemented a Denoising Diffusion Probabilistic Model (DDPM) to generate high-quality images with custom beta schedules.

Face Morphing

Face Morphing

Used Delaunay triangulation, affine warping, and image blending to morph between facial landmarks and generate caricatures.

Prokudin-Gorskii

Colorizing Prokudin-Gorskii

Reconstructed color photos from century-old negatives using SSD alignment, image pyramids, and edge detection.

Filters and Frequencies

Filters and Frequencies

Created hybrid images by constructing and manipulating Laplacian and Gaussian stacks across spatial frequency bands.