Lead AI Engineer · London
Abd Bastola.
Fetching the engine
Building the campus
The interactive campus is unavailable here
This experience needs WebGL and the 3D module. Your browser, graphics driver or connection may have interrupted it.
Nothing has been lost: this page, its URL and its content are intact. Everything the site contains — the CV, every article and case study, the repositories and the contact options — is ordinary readable HTML inside the world's own panels.
Notes on building production-ready AI agents, RAG systems, evaluation and enterprise AI delivery.
Prompt injection, data leakage, tool abuse, model supply chains and token-burning denial of service. A practical mental model for thinking about AI security the way you already think about the rest of your architecture.
A concise daily briefing on enterprise AI agents, RAG, evaluation, security and the ideas worth understanding before they become industry noise.
A confirmation tab will open — finish there, then check your inbox. One concise daily briefing, no spam, unsubscribe anytime.
6 articles
OpenAI's Astra launch is being framed as the start of something huge. A practical read on what the coverage actually signals about real-time multimodal and agentic models — transparency, control, safety, evaluation, cost and governance.
Voice agents are suddenly ready for the enterprise — which means teams are choosing between cascaded pipelines, all-in-one platforms and direct provider stacks. A practical look at the latency trade-offs, the compliance realities and the business impact that should drive the decision.
Business problem definition, data readiness, risk classification, model selection, MVP scope and measurable success criteria — the six questions that prevent model-first projects from becoming expensive regrets.
Document quality, chunking, metadata, retrieval evaluation, citations, access control and freshness — the unglamorous work that decides whether a RAG system actually helps people or quietly misleads them.
Evaluation, tool permissions, observability, failure handling, human approval and cost — the boring, unglamorous work that turns a promising agent demo into a system you can trust in production.
Independent writing on practical enterprise AI systems — no vendor spin, no hype, no “AI will change everything”. Just what works in production, and what quietly doesn’t.
If these notes help you think more clearly about AI, you can support the writing with a coffee.
