April 22, 2026
LAB-BAWUAH
Hi, I'm Brian. I build AI products where engineering, design, and the user meet.
I help teams turn AI from a demo into something the person on the other end can actually use, day after day. Human-centred thinking in every layer: search and reasoning, vision and documents, secure deployment, and the interface on top.
How I work
The best digital products emerge at the intersection of engineering, design, and the user.
Most AI products get built layer by layer, each layer optimised in isolation. By the time the user side is considered, the rest is locked in. End-to-end engineering keeps three concerns alive at once: what's actually buildable, what's usable, and what the person on the other end needs. No silo gets to decide for the others.
Live demo
Find the moment you remember.
A working agent over Offloading Thoughts, Mary Merz's travel podcast. Multilingual transcript search, geocoded entities, acoustic event detection, and an inline map, all driven by tool calling. Try it below.
AI can make mistakes. Always check the source.
Claude Haiku · Elasticsearch · IBM Granite Embedding · Azure Speech · Mapbox · Spotify
What I build
Four disciplines that compound in one engagement.
Search and AI orchestration. Vision and document intelligence. Secure deployment. The interface on top.
Search, retrieval & AI orchestration
Designing systems that turn questions into answers. Hybrid retrieval that mixes keywords with meaning, agentic pipelines that orchestrate models and tools, and structured outputs the rest of your stack can rely on.
Computer vision & document intelligence
Pulling structure out of images, scans, and documents that were never designed to be machine-readable. OCR, layout analysis, and visual models that turn pages into something you can search and reason over.
Sovereign & secure systems
AI that runs where the data lives. On a phone, in an air-gapped network, behind a security perimeter. A modern AI stack with none of the round-trips to someone else's cloud, built to survive a serious security review.
Product engineering & interface design
Real products shipped to real users. Interfaces designed around the person on the other end, not the data model underneath. Engineering that holds up at scale and looks good doing it.
How I engage
Three ways to bring me in.
Engagements are scoped to your reality: how much you've already figured out, what you can hand over, and the constraints of the environment you're building in.
01
Embedded engineer
I join your team for a fixed-term engagement and ship inside your environment: code, deployments, security review. Best when you have a product to build and need a senior pair of hands who already speaks the stack.
02
Advisory & architecture
Shorter scope: I review your AI architecture, retrieval design, or product strategy and write a concrete recommendation you can build from. A useful checkpoint before locking in a direction.
03
Fixed-scope build
A bounded deliverable: a RAG prototype on your own corpus, an on-device LLM proof-of-concept, a frontend for an existing model. Quoted up front, shipped end-to-end.
Get in touch
Building AI that real people will actually use?
Whether it's enterprise search, an agentic system, an on-device app, or the interface on top, I'd like to hear about it. A short note about the problem and your timeline is the fastest way in.
Recent posts
View on LinkedIn01 / 05
Field notes
View all →Creatives are the new athletes
May 16, 2026Why human value is shifting from producing to giving meaning. AI is eating production. Our work moves up the stack: to judgment, taste, and meaning.
10 min readData enrichment is the foundation for the AI era
May 4, 2026How we built a data enrichment pipeline inside a high-assurance enterprise using Elasticsearch, Python, and Kestra. Clean, connected data is the real foundation for AI.
6 min readMCP is powerful, but it can quietly break your AI system
November 17, 2025The Model Context Protocol makes it easy to plug tools into LLMs. But the default integration pattern floods context windows, burns tokens, and quietly turns systems fragile and expensive.
6 min read