01
Consulting
I work with technical and business teams to understand the workflow, identify where AI is genuinely useful, and decide what is worth building.
Independent practice · London
Hi, I'm Alex.
I design and build LLM-powered systems, workflow engines, and internal tools for organisations with complex technical and operational requirements.
My background is in astrophysics, production software engineering, and applied AI.
My Background
I started in astrophysics, moved into production software engineering, and now specialise in applied AI systems.
- 01
Astrophysics
Built large-scale data workflows, simulations, and statistical models for observational astronomy.
Download PhD Thesis - 02
Software Engineering
Built production forecasting systems operating across more than 1M SKU/store combinations, including automated retraining.
- 03
AI Architecture
Design and build LLM-powered workflows, agents, and internal tools, from initial architecture through to deployment and handover.
Engagement
How I work
02
Implementation
I build the application, integrations, AI architecture, and infrastructure—from working prototype through to a production system.
03
Handover
I document the important decisions, train the people who will operate the system, and transfer ownership so your team can extend it without depending on me.
Case work
Selected Projects
Finance
Gresham House
- Python
- FastMCP
- Snowflake
- Azure
- Entra ID
- Next.js
Built an enterprise MCP platform giving ChatGPT and Claude governed access to company knowledge and reusable capabilities across approximately 400 staff.
The platform turns source material into LLM-oriented wiki pages and serves versioned skills, prompts, and Snowflake retrieval guidance. It runs on Azure Container Apps with Entra ID, Docker, and Azure DevOps pipelines.
Finance
Downing
- LangGraph
- Azure
- HITL
- Microsoft Teams
Downing's digital team needed a better way to triage requests arriving through Microsoft Teams and route each issue to the right person.
I built an Azure-hosted LangGraph pilot with human-in-the-loop routing through Teams, persistent state in PostgreSQL, and observability around the orchestration layer.
The pilot demonstrated faster handoff and clearer ownership of incoming requests.
Agency / Ops
Digital Speed
- Next.js
- Vercel AI SDK
- Slack
- MCP
- RAG
Built an internal AI assistant that gives teams access to operational knowledge and tools directly inside Slack.
The application used TypeScript and Next.js with an agentic tool loop, alongside a GitHub-driven ingestion pipeline that detected changed company documentation, re-chunked it, and refreshed the relevant embeddings.
Startups
Hyperhumans
- FastAPI
- OpenAI
- SQLAlchemy
- RBAC
Built the backend for Neuralead, a lead-intelligence product that automated lead discovery, enrichment, scoring, and API delivery.
The system combined multiple enrichment stages behind a FastAPI service, with role-based access and credit metering for commercial API usage.
Startups
W4lkies
- Next.js
- FastAPI
- PostgreSQL
- Docker
A grassroots local project and one of my favourites.
I built W4lkies as a full-stack operations platform for a dog-walking business, bringing bookings, customer and dog profiles, invoicing, and expenses into one application.
Unlike most of the work above, I could observe this system being used day to day and continuously improve the workflows around real operational feedback.
Academic research
Published · 2018Ultra-Red Galaxies
PhD thesis
Research into some of the earliest massive galaxies in the universe, combining submillimetre observations, large-scale data reduction, statistical analysis, and scientific software.
Download thesisIndependent writing
Draft manuscriptFrom Seashells to Satoshis
A book about money and Bitcoin
An accessible exploration of how money evolved—from physical commodities and institutional ledgers to digitally scarce assets and Bitcoin.
Download draft