Senior engineer with 15+ years building low-latency, mission-critical front-office trading systems at Tier-1 investment banks (RFQ/OMS/FIX across Rates/FX/Equities). I focus on latency-sensitive flows, resiliency, and production ownership—where correctness, operational discipline, and rapid incident response directly matter.

I’m now extending that same engineering mindset into LLM-powered diagnostics and agentic workflows: systems that can turn noisy production signals into structured, evidence-backed explanations and next actions—while staying auditable and safe for real environments.

I am also currently completing the Nebius Academy – AI Performance Engineering course, a hands-on program focused on building and optimizing real-world LLM systems. The training covers core machine learning fundamentals and industry best practices, alongside a cohort of bright, highly motivated people from diverse technical and professional backgrounds. Key modules include LLM architectures, the transition from AI models to AI agents, ML Ops, performance engineering, and LLM post-training. You can explore my hands-on work in this space in my agent-building repository: https://github.com/sovereignagents/sovereign-agent-lab, and I will also be adding a dedicated repository showcasing neural network experiments, including building models from scratch in PyTorch. More details: https://academy.nebius.com/ai-engineering-uk

I use this GitHub space as a practical lab for production-grade AI in engineering workflows—especially reliability, incident investigation, and developer productivity.

• Smart Enterprise Diagnostics (SED): Open-source experiments in LLM/agentic incident investigation — timeline reconstruction, evidence-linked summaries, hypothesis ranking, and faster triage.
• AI Tooling Workflow: Experimenting with agentic IDEs and coding agents (e.g., Antigravity, Cursor, Codex) to stress-test where they help (and where they break) in real codebases and robust Java systems.
• Verifiable Automation: Moving beyond “black box” outputs — designing workflows with traceability, explicit assumptions, and reproducible steps.

• Core: Java (expert), concurrency/multi-threading, low-latency patterns, performance tuning, production debugging.
• Trading Domain: RFQ engines, OMS/EMS, FIX, ION MarketView, front-office integrations (Rates/FX/Equities).
• AI / Reliability: LLM orchestration, agentic workflows, evaluation/guardrails, prompt+tool design for engineering systems.

I’m interested in AI that improves outcomes under production constraints: measurable speed-ups in diagnosis, fewer repeated incidents, and better operational clarity—without sacrificing reliability.

Note: These projects are personal experiments, independent of my professional employment.