25+ years building, securing, and scaling enterprise systems. Author of a 65,000+ line production-grade agentic platform with Temporal workflow orchestration, hybrid RAG, MCP tooling, and cryptographic audit trails. I bring the same engineering rigour to every engagement.
I've scaled Fortune 500 network backbones, maintained 40Gbps global streaming infrastructure at 99.99% availability, delivered £325K annual AWS cost reductions in FCA-regulated fintech, and built 24x7 operations teams from the ground up.
I designed and built ACE — a 40+ microservice agentic collaboration platform running Temporal workflows, LangGraph orchestration, hybrid vector/BM25 retrieval, and a custom governance layer for AI agent operations. 65,000+ lines of production-grade async Python, not prototypes.
Production deployment of AI agents with Temporal workflow orchestration, MCP tool servers, and multi-provider LLM integration (Claude, GPT-4). HITL controls and governance boundaries built in from day one.
Hybrid retrieval pipelines: pgvector HNSW + BM25 full-text with RRF fusion and cross-encoder reranking. Document ingestion via Celery task queues, embedding with sentence-transformers, and knowledge graphs (Apache AGE).
Design and build internal platforms. 40+ microservice architectures on Docker with NATS JetStream event bus, asyncpg connection pooling, Redis caching, and full observability (Prometheus, Grafana, Loki, OpenTelemetry).
Reliability engineering for production systems. Circuit breakers, dead-letter queues, incident management, SLO frameworks, capacity planning, and on-call process design. £325K annual AWS cost reduction at YouLend.
Cryptographic audit trails with SHA-256 hash chains and Merkle roots. mTLS with CA lifecycle management, Ed25519 signing, JWT with JTI revocation, and append-only ledgers enforced at the database layer.
Enterprise network architecture, cloud migration, hybrid environments. BGP, peering, transit, zero-trust, and 25+ years of deep Linux systems expertise from Fortune 500 backbone to fintech production.
65,000+ lines of async Python across 40+ microservices. Not a weekend project — a production-shaped agentic platform with real security, real observability, and real governance.
Production Temporal deployments. Long-running workflows with deterministic event replay, failure recovery, and observable execution history. Not a cron job — a state machine.
pgvector HNSW + PostgreSQL BM25 full-text search, fused via Reciprocal Rank Fusion. Cross-encoder reranking (BAAI/bge-reranker-v2-m3). Freshness and authority boosting. Sub-second retrieval.
Full Model Context Protocol server with SSE transport. Tools for document search, SME index, AIOps knowledge base, agent control, and ledger queries. JSON-RPC 2.0 compliant.
Append-only PostgreSQL ledger with SHA-256 hash chains and Merkle roots. DB-level DELETE/TRUNCATE denial. Circuit breakers on NATS publish. Dead-letter queue with retry tracking.
Team topologies (parallel, pipeline, coordinator-worker, mesh), loop execution with stuck detection, and multi-tier capability assessment.
LangGraph StateGraph for document lifecycle orchestration. Celery distributed task queues for 5-stage ingestion with dead-letter handling and exponential backoff retry.
United Kingdom. Available for remote engagements globally and on-site within the UK.