I'm a systems-focused software engineer building compilers, low-latency engines, and performance-critical infrastructure in Rust and C++.
- 🎓 Computer Science graduate from University of York (July 2025)
- 💼 Seeking hybrid software roles in London focused on systems programming, infrastructure, and AI/ML engineering
- 🎯 Core interests: Compilers & GPU runtimes, fintech infra, AI tooling & multi-agent systems, networks & concurrent systems
- 🔭 Long-term vision: Become a specialist in high-performance systems, compiler optimization, and low-latency infrastructure
- 🧵 Compilers & GPU optimization — kernel fusion, graph rewriting, ONNX/TorchScript compilation
- ⚡ Lock-free systems & trading engines — atomic operations, order book matching, market microstructure
- 🧬 AI infrastructure & neuroevolution — evolutionary algorithms, topology mutation and hybrid optimization paradigms
- 🌐 Network protocols & concurrency — HTTP servers, async runtimes, TCP/UDP socket programming
- 🤖 Multi-agent AI systems — LLM orchestration, debate frameworks, knowledge synthesis
- 🖼️ Local-first applications — embedded ML, privacy-preserving architecture, semantic search
- 💹 Trading infrastructure — market data handlers, pre-trade risk, deterministic replay
| 🔧 Project | 🧠 Tech Stack | 📄 Summary |
|---|---|---|
| Tectra | C++, FlatBuffers, Prometheus |
Production-style trading infrastructure combining market data feed handler, pre-trade risk engine, strategy execution framework, backtesting engine, and deterministic replay with kill-switch circuit breakers |
| Image Browser | Rust, Tauri, React, SQLite, ONNX Runtime |
Local-first Pinterest-style image manager with semantic search via CLIP embeddings, visual similarity recommendations, and completely offline architecture with full user privacy |
| Vynapse | Rust |
Hybrid neuroevolution and deep learning runtime combining NEAT topology mutation, DEAP evolutionary search, PyTorch-style autodiff, and TensorFlow-style graph compilation into unified framework |
| AsteroidsAI | Python, Arcade, NEAT, DEAP, PyTorch, TensorFlow |
Real-time AI benchmarking platform comparing neuroevolution, genetic algorithms, genetic programming, evolution strategies, and graph-based RL on continuous-control tasks with emergent behaviour analysis |
| Chrona | C++ |
Git-inspired version control system built from first principles to understand content-addressed storage, immutable snapshots, commit DAGs, staging semantics, and distributed system fundamentals |
| Xyntra | Rust |
ML graph fusion compiler pass using e-graph rewriting to identify kernel fusion opportunities and emit optimized WGSL/PTX kernels for ONNX and TorchScript models |
| Nyquestro | Rust |
Lock-free limit order book matching engine exploring ultra-low-latency market microstructure design through atomic operations and safe concurrent data structures |
| Zyphos | Rust |
Network protocol laboratory implementing HTTP servers from scratch through 30 structured milestones, progressing from raw TCP sockets to HTTP/2, WebSockets, and QUIC |
| Debate CLI | Python, Typer |
Multi-LLM debate orchestrator with recursive summarization, async provider integration, and research-grade analytics for studying AI alignment and consensus formation |
- 🧪 Compiler optimization research — Writing technical paper on kernel fusion design patterns and e-graph rewriting strategies
- ⚡ Lock-free systems — Benchmarking atomic operations and designing wait-free data structures for trading engines
- 📊 Trading infrastructure — Building deterministic replay systems, pre-trade risk controls, and microsecond-latency order matching
- 🤖 Multi-agent AI frameworks — Orchestrating LLMs to surface emergent consensus and quantify model alignment
- 🖼️ Local-first ML applications — Embedding CLIP models for privacy-preserving semantic search without cloud dependencies
- 🌐 Network protocol implementation — Learning TCP state machines, event loops, zero-copy I/O, and kernel bypass techniques
- 🧬 Neuroevolution & comparative AI — Benchmarking how evolutionary vs gradient-based approaches solve identical continuous-control problems under sparse rewards