B.Sc. Computer Science · University of Southern California, 2027
CS + ML research + robotics + aerospace systems engineering. I build things that need to work under real constraints — distribution shift, sparse data, deployment-limited compute, and permanently shadowed lunar craters.
Ten independent, reproducible research frameworks — ~33,000 lines of Python, ~5,300 lines of tests, 310 pytest cases. Each module implements synthetic ground-truth DGPs, YAML-configured benchmarks, and structured evaluation reports. The unifying question: how do modern ML estimators degrade as their identifying assumptions are systematically violated?
| # | Module | Description | Repo |
|---|---|---|---|
| 01 | Causal ML | Meta-learners (S/T/X/R/Causal Forest), TMLE, NOTEARS, PCMCI, off-policy evaluation | causal-ml-decision-making |
| 02 | Foundation Model Adaptation | LoRA, adapter bottlenecks, prefix tuning, influence functions (LiSSA, TracIn), calibration | foundation-model-adaptation |
| 03 | Reinforcement Learning | CQL, IQL offline RL; CMDP Lagrangian safe RL; ensemble world models; CEM planning | reinforcement-learning-real-environments |
| 04 | Bayesian Deep Learning | Variational BNNs, MC Dropout, deep ensembles, conformal prediction (RAPS), calibration | probabilistic-bayesian-deep-learning |
| 05 | Generative Modeling | DDPM/DDIM, classifier-free guidance, flow matching, diffusion posterior sampling | generative-modeling-scientific |
| 06 | Graph ML | GCN, GAT, GIN, GraphSAGE, DiffPool, temporal GNNs, link prediction | graph-ml-relational-systems |
| 07 | Self-Supervised Learning | SimCLR, MoCo, BYOL, VICReg, Barlow Twins, MAE; linear probe + k-NN eval | self-supervised-representation-learning |
| 08 | Neurosymbolic Reasoning | Differentiable forward chaining, neural theorem prover, typed program synthesis | neurosymbolic-reasoning |
| 09 | Federated & Privacy ML | FedAvg/FedProx/FedNova, DP-SGD, Rényi DP accounting, pFedMe, secure aggregation | federated-privacy-preserving-ml |
| 10 | Scientific ML | PINNs, E(3)-equivariant GNNs, scVAE, FNO, Bayesian optimization for materials | ml-for-science |
| Project | Description | Repo |
|---|---|---|
| SpaAIder | Local-first multi-agent agentic runtime — TypeScript orchestration plane, Python FastAPI memory substrate, 135-agent taxonomy, episodic-semantic-vector memory, dependency-aware parallel scheduling | SpaAIder |
| URC Autonomy Stack | ROS 2 (Humble) full autonomy for University Rover Challenge 2026 — 7-target mission orchestrator, EKF sensor fusion, GNSS waypoint nav, ArUco visual servoing, YOLO detection | urc-cv |
| Curriculum Vitae | LaTeX CV (auto-compiled, 13 pages) | curriculum-vitae |
This is a curated list of my featured work. GitHub limits pinned repos to 6, so this profile serves as the full directory.