Head of Engineering Automation at Cortekz Technologies (UK). I build AI systems for engineering-heavy industries - Oil & Gas, Infrastructure, and Chemical Processing - where the problem space involves structured documents, complex data relationships, and workflows that have historically resisted automation.

My core work sits at the intersection of LLM pipelines, knowledge graph engineering, and applied NLP on industrial data. I also ship independent products in HR-tech and edtech, which keeps the full-stack and product side sharp.

• P&ID Digitisation - parsing and extracting structured data from Process & Instrumentation Diagrams using computer vision and NLP pipelines
• DEXPI/XML Processing - ingesting and transforming DEXPI-standard engineering data using pyDEXPI; building compliant data models for downstream automation
• SDC/IDC Compliance - automated checking of engineering deliverables against discipline-specific compliance rules using LLM-backed agents
• Enquiry-to-Proposal Automation - end-to-end agentic pipeline for processing RFQs, vendor bid evaluation, and proposal generation

• Neo4j - modelling P&ID topology, equipment hierarchies, and tag relationships as graph structures for traversal and reasoning
• Graph ingestion from DEXPI XML; custom Cypher query layers for compliance and cross-discipline checks
• Graph-backed retrieval as an alternative to pure vector RAG for structured engineering data

• LangChain - chain and agent construction for multi-step document workflows; tool use, memory, and retrieval integration
• Agentic AI - multi-agent orchestration for document-heavy processes: contract review, compliance checking, bid evaluation, enquiry processing
• vLLM - self-hosted model serving for local inference; deploying and optimising open-source LLMs for cost-controlled production use
• Retrieval-Augmented Generation over engineering corpora - combining graph context with vector search
• Prompt engineering and output structuring for domain-specific extraction tasks

• Azure - VM deployment, Azure Pipelines for CI/CD, cloud resource management for AI workloads
• Docker - containerised deployment of AI services and FastAPI backends; reproducible environments across dev and production
• Windows Server / WSL2 production environments for on-premise client deployments
• FastAPI backends; REST API design for AI-assisted engineering tools
• Python-first stack with strong emphasis on pipeline reliability and traceability

Most of the meaningful work is proprietary - industrial AI systems don't ship to public repos. This profile surfaces the public-facing slice: applied AI prototypes, voice and retrieval interfaces, and product builds. The engineering depth behind it is in production at client sites across the UK and South Asia.