Mostly interested in reinforcement learning, numerical linear algebra, and ML systems. I enjoy working on problems that involve both mathematics and programming.
Started at Tenstorrent - low-level C/C++ ML kernels for AI accelerators
Finalist at the UCL Full House Poker Bot Tournament
Codeforces Candidate Master unlocked. Turns out purple is definitely my color
Built a platform to simulate funding for overlooked humanitarian crises and ended up taking 3rd at the ETH Zurich Datathon
Snagged a Distinction @ Potyczki Algorytmiczne
A simulation environment for training and benchmarking multi-agent reinforcement learning on procedurally generated enterprise and OT networks, featuring partial observability, durative actions, and synthetic SIEM telemetry.
DRF-based allocation of heterogeneous CPU/GPU resources under fairness and capacity constraints.
ML Systems · Robotics
C++23 real-time inference runtime for robotics. Static memory, deadline-aware scheduling, shared-memory IPC, Mamba and flow-matching support. Deterministic latency benchmarks with p50/p95/p99.
ML Inference · Systems
Persistent inference session management for NVIDIA Triton. Suspend / Resume / Restore with checkpointing, state serialization, crash recovery, and async I/O. Working PoC with ONNX Runtime CPU backend.
MARL · Cyber Defense
Multi-agent cyber defense environment - event-driven, partially observable, graph-based observations. Asynchronous agent interactions, adversarial learning, reproducible RL evaluation infrastructure.