How the problem was framed, what was decided, and what was built.

The problem

Integrating AI into an existing education product, it was unsettled which features should be delivered by an LLM and how they would be operated.

Scope

• •Requirements and design for LLM features
• •Implementation within a three-person core team under the CTO
• •Building the evaluation flow

Outcome

The LLM features went live as a supported part of the product, with an evaluation procedure in place to keep improving them.

The problem

The web platform behind a healthcare AI product needed work, and internally, finding information and answering questions was slow.

Scope

• •Leading web development and optimisation for the platform
• •Designing and building a RAG Slackbot over internal knowledge
• •Forming and running a three-person engineering team

Outcome

RAG-based internal knowledge search entered daily use, and optimisation of both frontend and backend of the platform was completed.

The problem

Launching a new platform meant deciding architecture, development process, and deployment from nothing.

Scope

• •Leading engineering as CTO
• •Architecture, CI/CD, and deployment
• •Managing a six-person engineering team

Outcome

Everything from architecture to deployment was put in place, and a six-person team was able to develop continuously.

The problem

Core backend services for Layer 2 infrastructure had to go from design to implementation in a short window.

Scope

• •Design of the core backend services
• •Implementation in Go

Outcome

Within a short engagement, the core backend services were designed, implemented, and handed over.

The problem

On a platform handling freehand input, drawing quality and PDF export both needed work.

Scope

• •Canvas rendering optimisation
• •Stroke interpolation algorithms
• •Improving the PDF export pipeline

Outcome

Drawing quality and export processing improved; the role also covered data science work within an English-speaking team.

The problem

Manual listing work had become a constraint on the business, and they needed someone to own automation from design through build.

Scope

• •Architecture for the whole automation system
• •Browser automation, admin console, and data pipelines

Outcome

End-to-end listing automation went into operation and was designed and run single-handedly over three years.