Brighton Web Development Meetup

Tonight we are delighted to be having two talks, the first from Andrew Cheeseman and then the second talk by George Oxley and Seb Davies

TALK 1:

From Zero to AI
How a Small Studio Rebuilt the Way It Ships Software - Andrew Cheeseman

Tonight, we are delighted to host longtime group member Andrew Cheesman's new talk on his experiences of implementing AI into software development projects that Andrew worked on with his team over the last couple of years.

TALK SYNOPSIS

Everyone’s talking about AI — but what does adopting it actually look like in a small company? Not a research lab, not a big-tech giant, but a handful of people shipping real software for real clients.

In this talk, Andrew Cheeseman shares the honest, unpolished story of how his studio went from using no AI at all to weaving it through the whole delivery process: writing and reviewing code, running DevOps and SRE, and even how the business communicates with and delivers for its clients — including a Fortune 500 account.

Expect the good, the awkward and the genuinely surprising: what worked, what quietly wasted time, where the team pushed back, and the judgement calls that mattered far more than any particular tool.

If you run or work in a small team and you’re wondering how to make AI a practical multiplier rather than a hype cycle, this session is for you.

WHAT YOU’LL TAKE AWAY

• A realistic path from zero to embedded— no big budget, no big team required.
• Where AI genuinely earns its keep across coding, DevOps/SRE and client delivery — and where it doesn’t.
• How to bring a sceptical team along while keeping quality and review discipline intact.
• The pitfalls to sidestep, from someone who walked into a few of them.

++++++++++++++++++++++++

TALK 2:

From Prediction to Power: Building an AI Revenue Optimiser for a Solar Plant

How do you turn machine learning predictions into safe, valuable operational decisions in a real power plant?

George Oxley and Seb Davies, software engineers at Brighton-based DabApps, will share how the team built a revenue optimiser for a concentrated solar power (CSP) plant in South Africa using Python and PyTorch.

At its core is a custom LSTM digital twin that predicts how the plant will behave, combined with an optimiser that searches for ways to operate the plant more effectively.

But predicting what might happen was only part of the problem. The much harder challenge was making sure the optimiser’s recommendations actually made sense in the physical world.

George and Seb will explore how they introduced real-world constraints around ramping, thermal storage and dispatch, worked with domain experts to refine the model, and tackled messy problems including unreliable sensor data, feature engineering and look-ahead bias.

It’s a practical look at what happens when machine learning leaves the prototype stage and has to make useful decisions within the constraints of a complex real-world system.

SPEAKER BIOS: