Tom Stacey
What I Build

I build AI products that are honest about what they are.

Four live tools: a digital twin for a brewery, a dissertation coach for students, a relationship mediator and a university module on trust. Plus over a decade in the EV world, and the engineering underneath all of it.

Funded · Live pilot · Spinout track

BrewAI

Digital twins of working breweries that show brewers where the energy goes, and how to cut it.

What it is

BrewAI turns a working brewery into a live digital twin. Circuit-level sensor and energy data, captured continuously for nearly three years, feeds AI models that let brewers simulate changes, cut energy and carbon, and produce verifiable emissions data they can actually stand behind.

The interesting bit

Most sustainability AI runs on estimates and annual averages. BrewAI has almost three years of continuous, circuit-level energy data from a live brewery, down to individual pieces of equipment. There's no shortcut to data like that, you have to collect it day by day.

Role

Co-founder and Technical Lead. Designed the architecture, data pipeline, models and product from first prototype onward.

Proof points
  • Backed by Innovate UK's iCURe programme (Stage 3) and EIT
  • ~3 years of continuous circuit-level sensor and energy data
  • Live brewery pilots UK-wide
  • On a spinout track
Talk to me about pilots or partnership
BrewAI Energy Intelligence forecast showing 71% variance explained, 14.4% out-of-sample MAPE, RMSE 13.22 kWh per day, trained on 91 days
Energy intelligence - the 14-day forecast, warts and all (worst fold: 38.1% MAPE)
Ask BrewAI natural-language query surface with energy cost, load, carbon and data-coverage metric cards
Ask BrewAI - natural-language queries over the brewery's real data
MVP · In discussion with HEIs

Ship My Dissertation

Project management built for the one big project most undergraduates have never done before.

What it is

Ship My Dissertation gives students a plan, milestones and AI-guided support for the undergraduate dissertation, a process that usually runs on panic and a January all-nighter. It treats the dissertation as what it is: a real project most students have never been taught to run. It's a working MVP, and I'm now talking to universities about pilots.

The interesting bit

Fifteen years of teaching project management went into this, and I wrote the code as well. I didn't have to describe the tool to a developer and hope, I could just build it.

Role

Founder and Builder.

Proof points
  • Working MVP, live and usable
  • Active implementation discussions with HEIs
  • AI-guided support built in
In pilot discussions with universities. If that's you, let's talk.
AI Transparency Log confirming AI was used exclusively for project planning and time management, not for generating academic content
The AI Transparency Log - the AI helped plan the project, never wrote a word
Weekly check-in form with a mood selector ranging from Struggling to Smashing it
Weekly Check-in - self reporting that the coaching is built on
Closed Beta · Self-hosted core

Harmony Mediator

A privacy-first AI relationship mediator. The intimate conversations stay on a self-hosted model.

What it is

Each partner talks privately to a neutral AI mediator, which works out what matters to the other person and returns separate guidance to each side. Neither party ever sees the other's raw words, a caucus model borrowed from professional mediation. The mediator core runs on a self-hosted, open-weight LLM so intimate conversations never leave home turf, with frontier models reserved only for specific safety-critical roles.

The interesting bit

Anything intimate stays on a local, open-weight model. Frontier models come in role by role, only where safety genuinely needs them. And the caucus design means there's just no route for one partner's words to reach the other.

Role

Founder and Architect.

Proof points
  • Caucus model: neither partner ever sees the other's actual words
  • Mediator core runs on a local, open-weight LLM
  • Frontier models only for specific safety-critical roles
  • Encrypted in transit and at rest · single trusted operator · minimal retention
I'm interested in conversations about safety, ethics and commercialisation
Harmony Mediator architecture diagram showing the caucus model: two private conversations synthesised by one sovereignty core
The caucus model - two private conversations, one mediator in the middle
In Delivery (Year 2) · E-learning next

Who Can You Trust?

An interdisciplinary module on working out who to trust, now that deepfakes, scam reviews and AI slop are everywhere.

What it is

An eleven-week online module about deciding what to believe, when a deepfake, a five-star scam review and a friend's hot take can all land in your feed inside five minutes. Fully online, fully interdisciplinary, open to any student on any course, with no exam: students build a real analysis of a manipulation case and reflect on it. Now in its second year at ARU, and being packaged as standalone e-learning.

The interesting bit

Because it's fully online and open to any course, any student at an institution can take it, and there's no exam to timetable: assessment is a real manipulation-case analysis instead. It's been through one full cohort at ARU already, with the student feedback to show for it.

Role

Designer and Lecturer.

Proof points
  • Eleven-week module, in its second year at ARU
  • Fully online, interdisciplinary, open to any course
  • Authentic assessment (manipulation-case analysis plus reflective essay), no exam
  • E-learning packaging planned
“I went in half-assed, convinced I knew how to sniff out the lie already. I was wrong.”
Student, “Trust Journey” reflective essay (anonymised)
Packaging this as standalone e-learning. Interested institutions welcome.
Lecture slide: Seven levers that open the System 1 doors, mapping Cialdini's principles
Week 3 - “Seven levers that open the System 1 doors”
Lecture slide: Trust you don't deserve, the acquisition protocol, with the Know / Like / Trust framework
Week 3 - “Trust you don't deserve: the acquisition protocol”

Credibility

A selection of the things I do and who I've worked with…

30+
BBC TV & radio appearances (automotive and EV)
6 yrs
Vice President, Tesla Owners UK
12 yrs
EV driver since 2014
10+ yrs
Senior Lecturer, Anglia Ruskin University
MSc
Computer Science · FHEA · MBCS
2012
STEM Ambassador since · Essex County STEM Award 2013
2022-2026
Speaker: Everything Electric Amsterdam · Supercharged UK
Writer
The Conversation

On Teaching

“I emailed him the other night with questions about the assignment and he responded within hours, which is really a massive thing at this university. It is often difficult to contact staff for help. This is a massive deal for me personally.”
BSc student, ARU, 2025
“Tom is very engaging and clearly passionate about what he teaches.”
MSc student, ARU, 2025
“Tom has been a great module leader, he makes the classes extremely engaging and actually enjoyable to join.”
BSc student, ARU, 2025
About

I build AI products, lecture at Anglia Ruskin University, and have spent over a decade in the EV and automotive world.

All of it is technology that has to work in the real world: a brewery's carbon data, a student's first big project, the future of electric transport. MSc in Computer Science, Fellow of the HEA, Member of the BCS.

Download CV