Shape the system
Turn ambiguous business problems into reusable AI platforms, blueprints and patterns with clear boundaries, ownership and operating models.
I design frontier AI systems, enterprise platforms and intelligent automation that move from an ambitious idea to dependable production.

The interesting part is not making AI respond. It is shaping the data, tools, guardrails and interfaces around it until the whole system becomes dependable.
PRODUCTS · FRAMEWORKS · EXPERIMENTS
I work between what AI could do and what an organisation can trust it to do every day.
Turn ambiguous business problems into reusable AI platforms, blueprints and patterns with clear boundaries, ownership and operating models.
Design tools, context, evaluations and guardrails around agentic and multi-agent workflows so prototypes can operate responsibly in production.
Work with business leaders, engineers, product teams and data professionals to make advanced capabilities understandable, adoptable and measurable.
Prototype emerging technologies, translate research into practical systems, mentor teams and share the reasoning behind architectural decisions.
Working across: agentic AI, multi-agent systems, machine learning, Model Context Protocol, platform engineering, enterprise architecture and intelligent automation.
The standard: secure, reusable and production-ready solutions that create measurable business impact without hiding complexity behind theatre.
I lead the design and delivery of a frontier internal AI platform inspired by cowork-style workflows, bringing agentic AI directly into everyday user workflows. I also contribute to enterprise-wide AI blueprints and reusable architecture patterns that accelerate adoption, consistency and responsible scale.
Used the Model Context Protocol and LLMs to move enterprise applications from prototype to production.
Built ML and DL models from historical test-case error patterns, evolved the solution into an LLM response chatbot and prototyped automated API test generation from OpenAPI specifications.
Completed the advanced programme in machine learning and deep learning. View credential ↗
Contributed to the test automation framework now open-sourced as CafeX, while creating chatbot prototypes and supporting NLP pipelines.
Classified around 7.5 million metadata attributes into Protection Groups using machine learning and a custom combination of text-classification strategies.
Worked across simulation data management, sensor analytics, Predix operations, asset performance management and ServiceNow IoT alerting workflows.
The starting point for a way of thinking that still shapes how I approach constraints, architecture and production.
Notes on agents, tools, protocols and evaluations—the parts of AI architecture that become visible only when you try to ship.
Why the shape of a tool—and not just the model calling it—changes the reliability, cost and behaviour of an agentic system.
READ THE ESSAY ↗