I work on machine learning research: quantum ML and quantum error correction, robustness and safety of AI agents, and small efficient models. I also build the tools I use every day.
Every research repo ships a LIMITATIONS.md and pre-registered hypotheses, so the numbers mean what they say.
| Project | What it asks |
|---|---|
| SurgeryNet | Can a learned decoder handle lattice surgery? A GNN decoder with learned-decoder accuracy on merge/split operations, including timelike errors. |
| EntangleRSA | What does entanglement buy a quantum RL policy? Compares PQC and MLP PPO policies with representational similarity analysis across entangling depths. |
| DequantEngine | Given a trained quantum classifier, builds the closest classical tensor-network surrogate and reports where it catches up. |
| Mirage | A controlled adversarial web for measuring how browser agents fail under prompt injection. |
| Project | What it is |
|---|---|
| Aaवाz | 2.4M-parameter text-to-speech for 12 Indian languages that runs on CPU with only NumPy. |
| PRABHAT | Extreme-weather anomaly tracking through ensemble forecasts (Smart India Hackathon, SIH26078). |
| VISHWAS | Byzantine-fault-tolerant integrity layer for drone swarms. |
| Lex | Planner, coder and reviewer agents that work on a codebase together. |
LenovoRGB · QuietHours · GlassyTableau
Stack: Python, PyTorch, PennyLane, NumPy, TypeScript, Kotlin, C++