The Lab
Where I'm pointed — AI, and fast.
AI changes the math on what's worth building and how long it takes. This is what I can build with it, what I'm learning, and where it's heading.
Build with AI — fast
Here's the shift: with AI in the loop, things that used to take months now take weeks — sometimes days. Same engineering rigour, a fraction of the time. These are the AI builds I can turn around quickly.
Document intelligence
OCR and LLM pipelines that read invoices, claims or forms, extract structured data and validate it — the kind of system I lead in production today.
AI agents & assistants
Task-running agents and copilots wired into your tools and data, so they do real work, not just chat.
RAG over your data
Ask-your-documents search and Q&A grounded in your own knowledge base, with vector search (Qdrant) and citations.
Workflow automation
AI steps dropped into event-driven pipelines — classify, summarise, route and act — on top of solid queue-based architecture.
Chatbots & support
Customer-facing assistants that know your product, hand off cleanly, and stay on-brand.
Content & data pipelines
Generation, enrichment and transformation pipelines that turn messy inputs into clean, usable output at scale.
Timeframes are indicative and depend on scope and integrations — but the direction is the point: AI makes delivery dramatically faster.
What I'm learning
I'm deliberately pointing my backend and systems experience at AI. This is where I'm investing — and what you'll be hiring into as it matures.
LLM-powered backends
DeepeningDesigning backend systems where language models are first-class components — structured output, tool use, evaluation and guardrails.
Agentic automation
DeepeningMulti-step agents that plan, call tools and recover from failure — reliable enough to put in front of real business workflows.
Vector search & retrieval
Hands-onQdrant, embeddings and retrieval strategies that make AI answer from the right context, not hallucinate.
AI-native architecture
ApplyingEvent-driven, queue-based systems (RabbitMQ) with AI steps built in — scaling intelligent automation without the fragility.
Where this is going
The next wave of products will be AI-native from day one. My plan is simple: pair solid backend engineering with AI so small teams and founders can ship things that used to need a whole department — faster, cheaper and genuinely useful.
What's next
Where I'm investing next — concepts and R&D where AI, automation and good engineering meet. If your business lives in this space, let's build it together.
AI Wellbeing Shield
A Manifest-V3 Chrome extension that understands the meaning of a post — not just keywords — and gently blurs content that may harm mental wellbeing (health anxiety, doomscrolling, gambling). The user always stays in control: Show Anyway.
Hotel Outreach Engine
A pipeline that auto-generates a personalised one-page website for each small hotel — colours, photos and copy pulled from its own Google Maps presence — then packages a bilingual outreach message. Design-to-lead, automated.
LLM Lab
Ongoing experiments with retrieval, agents and vector search (Qdrant) — the groundwork for the AI-native products coming next.
Let's build something
worth shipping.
A project, a product, or a small team to deliver it — tell me what you need and I'll tell you how fast it can happen.
Start a project