London · Engineering & Applied AI

James Watson,
building AI systems
for hard problems.

A hands-on engineering lead who stays close to the code, across models, pipelines and infrastructure.

9yrs Software engineering
Full Stack
Cloud Infrastructure
AI Multi-Agent Systems
Machine Learning
Evaluations & Guardrails
MSc Artificial intelligence
Distinction
University of Leeds
BSc First Class (Honours)
Mathematics
King's College London

01 Profile

Equal parts builder and technical lead.

I lead engineering at CarbonAI, a London startup building AI-powered software at the intersection of climate data and carbon markets. The role is deliberately hands-on: I set technical direction and still write the code that ships; across LLM pipelines, multi-agent systems, retrieval, cloud architecture, data infrastructure and the React frontend people actually use.

My foundations are mathematical: a First in Mathematics from King's College London, then a Distinction MSc in AI from the University of Leeds. I like understanding systems all the way down, from how a transistor switches to how a fleet of agents coordinates. My strengths lie in my ability to learn, understand and then apply concepts in the technical domain whilst having the business acumen to understand product direction, operational efficiency and strategy.

role     Head of Engineering
company  CarbonAI
based    London, UK
domain   Climate · Carbon markets
modes    IC & leadership

02 What I work on

Four layers I move between daily.

From the model and agent layer down to the networking and data that keep it private, reliable and observable.

L1 / Applied LLM systems

Models & agents in production

Designing multi-agent architectures and the pipelines around them; retrieval, tool use and orchestration that serve thousands of users per day.

Multi-agentTool useRAGMCP serversAzure AI FoundryLLM pipelines
L2 / Cloud & platform

Secure, scalable infrastructure

Private networking done properly (endpoints, DNS and VNet integration) plus the auth, routing and scalability that make a platform safe to run.

AzureAWSPrivate endpointsVNet / DNSFront DoorAuth0App ServiceLoad BalancingAPIs
L3 / Data engineering

Pipelines & retrieval at the base

The data layer behind the product: transactional and analytical stores, vector and semantic search, and telemetry to see what's really happening.

SnowflakeCosmos DBAzure AI SearchPower BIKQL / App InsightsPython
L4 / Product & front end

Interfaces people use

The usable surface of the product: the CarbonAI chat experience, dashboards and a reporting portal for cost transparency and auditability.

ReactTypeScriptProduct designReporting

03 Selected work

My work, by the numbers.

Projects I've built, and the numbers behind them: from production AI to enterprise data infrastructure and research.

Platform · Applied AI CarbonAI · present

Multi-agent platform for carbon markets

5,000+ Users served
200k Documents indexed
90%+ Eval score across
500+ research problems

A multi-agent system over climate and carbon-market data, with retrieval, tool use and orchestration running in production. I architected, built and deployed it, with a continuous evaluation suite of 500+ research problems keeping quality in check.

Multi-agentRAGTool useEvaluationsAzure AI Foundry
Infrastructure · Cloud AWS · Azure

Secure, multi-cloud architecture

6+ yrs Cloud engineering
21 Environments
engineered
AWS + Azure Multi-cloud,
in production

Private networking done properly across both major clouds: endpoints, VNet/DNS integration, auth and routing that make a platform safe to run at scale.

Private endpointsVNet / DNSInfrastructure as CodeState MachinesLoad BalancingMicroservices
Data engineering CBRE · Balfour Beatty

Enterprise data pipelines at scale

50+ Pipelines built
Millions Of rows processed
Multi-TB Data under management

Production data infrastructure across the property and construction sectors, ingesting messy, real-world sources into dependable analytical and transactional stores.

PythonLambdaDAGETL / ELTAWSRedshift
Research · MSc University of Leeds

Image-based RAG with Vision-Language Models

84.4% Accuracy achieved
+35.5pts vs 48.9% naive
baseline
Distinction MSc research project

Distinction research project on image-based retrieval and Vision-Language Models, lifting accuracy from a 48.9% naive baseline to 84.4%.

VLMsImage-based RAGComputer visionEvaluation

04 Background

How I got here.

From mathematics into data, and from data into applied AI and engineering leadership.

01

BSc Mathematics — First Class (Honours)

King's College London

The mathematical grounding that still shapes how I reason about systems, logic and uncertainty. In-depth training in calculus and linear algebra has proved particularly useful.

02

Data Engineering - Junior to Senior

CBRE · Balfour Beatty

Earlier roles focussed on data pipelines and infrastructure across the property and construction sectors. This is where I learned to make messy real-world data dependable. I also accrued deep experience in enterprise Cloud Infrastructure (primarily AWS) and gained my passion for building things. Focus shifted towards AI and ML systems during this time.

03

MSc Artificial Intelligence — Distinction

University of Leeds

A move deeper into AI and Machine Learning, formalising the foundations I'd been building toward. I developed a broad theoretical and practical understanding of topics ranging from Robotics and RL to Deep Learning, Computer Vision and LLMs. Completed a research project with Distinction, focused on Image-based RAG and Vision Language Models (VLMs).

04

Head of Engineering

CarbonAI · London · present

I joined CarbonAI as a Senior AI/ML Engineer but quickly showed my worth and leadership abilities. Now I am leading engineering for an AI product in climate and carbon markets, staying close to the code while setting the technical direction and strategy.

05 Beyond the terminal

The other open tabs.

Threads I keep pulling on outside of work.

/ markets

Investing

I enjoy learning about economics, finance and market mechanics. A recent book I enjoyed: Chokepoints by Edward Fishman.

/ pigment

Painting

The meditative process of creating something on canvas is uniquely special. I particularly enjoy Caravaggio's work.

/ fitness

Strength & Sport

I'm an avid Gym-goer and play racket-sports weekly. Currently Squash and Badminton are the favourites.

/ origins

History & etymology

A deep curiosity about where ideas, institutions and words actually came from.

06 Get in touch

Open to interesting problems.

Whether it's applied AI, climate-tech engineering, or comparing notes on dividend strategy, I'm happy to hear from you.