Contributed to a research collaboration between Georgia Tech and the University of Florida, classifying lizard vertebrae fossils by species and spinal position using a neural shape model autoencoder. The pipeline encodes 3D vertebrae meshes into latent vectors, applies NCA metric learning, and runs supervised classifiers on the transformed embeddings.

Rigorous study of algorithm design paradigms: divide and conquer, dynamic programming, graph algorithms, NP-completeness, and linear programming. Heavy emphasis on proofs of correctness and runtime analysis. The course forced a level of mathematical precision that sharpened how I reason about code complexity.

From continuous control to multi-agent cooperation to real-world autonomous racing. Progressed through deep RL architectures (DDPG, PPO, MAPPO), learning that reward function design matters more than network architecture.

Focused on neural approaches to NLP, from text classification and language modeling to distributional semantics. Assignments were entirely code-driven using PyTorch -- building models from scratch rather than relying on pre-built frameworks.

Comprehensive tour of supervised, unsupervised, and reinforcement learning fundamentals. Each assignment was a deep-dive experiment comparing algorithms on real datasets -- not just implementing them, but understanding when and why they succeed or fail.

Covered the full breadth of computer vision, from low-level image processing to high-level recognition. Implemented algorithms for optical flow, frame interpolation, object tracking, stereo correspondence, and face recognition from scratch. Every assignment produced tangible, visual output -- making it easy to see exactly where an algorithm succeeds and where it falls apart.

Applied core AI techniques to simulated robotics problems: Kalman filters for tracking, PID controllers for flight stabilization, and search algorithms for warehouse navigation. Each project was a self-contained simulation where the algorithm had to operate under noisy sensors, uncertain dynamics, and real-time constraints.

Covered the full software development lifecycle from requirements to testing. The team project built an Android app for comparing job offers with customizable comparison weights. This course was less about individual algorithms and more about working effectively as a team on real software.

Hands-on course in data engineering and interactive visualization, culminating in a team project building a Spotify Music Recommender. Combined D3.js frontend development with large-scale data processing, graph algorithms, and API integration.

Foundational AI covering search, probabilistic reasoning, and temporal modeling. Implemented core algorithms from scratch -- A* on real road networks, Bayesian inference with Gibbs sampling, and Viterbi decoding for gesture recognition.

Systems programming in C covering threads, synchronization, inter-process communication, and distributed systems. Built multi-threaded servers, IPC mechanisms, and proxy caches. Working at the systems level gave me a much deeper appreciation for what happens beneath the abstractions I use daily.