Hello, I'm Ben π
Prefer a quick overview? Use the tabs to explore the highlights or head back to the terminal view.
Read a short introduction about me.
Hi there. My name's Ben, and I currently head up AI & Machine Learning at Motorway, leading teams that build applied AI products powering the UK's fastest-growing used vehicle marketplace. My work sits at the intersection of AI, product, and engineering β turning complex machine learning and AI into reliable, safe, and commercially impactful solutions.
In addition to my day job, I advise startups on AI, ML, and data science strategy β helping them design, build, and operationalise intelligent systems, and have spoken at a number of conferences including as a main stage speaker at Google Cloud's London Summit in 2024 and Big Data London in 2025.
Before Motorway, I led ML at computer vision startup DeGould and worked as a technical consultant for 4 years across Accenture, Anglo American, and the UK's Ministry of Defence. My consulting experiences allowed me to hone my ability to spot commercial opportunity β and I take pride in ensuring every AI initiative is grounded in adding real business or user value.
With a hands-on foundation in data science and ML engineering, my focus more recently has been on delivering transformational experiences with agentic generative AI systems. I'm passionate about building high-performing teams and creating ethical, scalable AI systems that drive real impact.
Want to know more? Just ask a question here, and my assistant will do its best to help!
In addition to my day job, I advise startups on AI, ML, and data science strategy β helping them design, build, and operationalise intelligent systems, and have spoken at a number of conferences including as a main stage speaker at Google Cloud's London Summit in 2024 and Big Data London in 2025.
Before Motorway, I led ML at computer vision startup DeGould and worked as a technical consultant for 4 years across Accenture, Anglo American, and the UK's Ministry of Defence. My consulting experiences allowed me to hone my ability to spot commercial opportunity β and I take pride in ensuring every AI initiative is grounded in adding real business or user value.
With a hands-on foundation in data science and ML engineering, my focus more recently has been on delivering transformational experiences with agentic generative AI systems. I'm passionate about building high-performing teams and creating ethical, scalable AI systems that drive real impact.
Want to know more? Just ask a question here, and my assistant will do its best to help!
Get my contact details and social links.
Get in touch:
Email btjones.me+contact@gmail.com
GitHub https://github.com/btjones-me
LinkedIn https://www.linkedin.com/in/benthomasjones/
Email btjones.me+contact@gmail.com
GitHub https://github.com/btjones-me
LinkedIn https://www.linkedin.com/in/benthomasjones/
Download my current CV as a PDF.
Learn how this app is built and how the AI works.
App architecture: A lightweight Flask app serving a terminal-style UI with an AI backend. It serves as a demonstration practicing context engineering, guardrails, rate limits, and logging/observability. Commands are handled through a registry and a summary view; the design is intentionally minimal but shows solid engineering hygiene, safety awareness, and configurability rather than just raw model calls.
AI implementation: An LLM service built with Pydantic AI agents, the Pydantic AI Gateway, Logfire, and Google Gemini (default gateway/google-vertex:gemini-2.5-flash, switchable to 2.5 Pro). It uses a hardened system prompt focused strictly on Benβs professional background plus a small knowledge base file, and Google safety settings. Guards enforce input validation (length, injection, repetition) and sanitize outputs. Sessions are tracked per session_id with history trimming based on configured turn limits.
AI implementation: An LLM service built with Pydantic AI agents, the Pydantic AI Gateway, Logfire, and Google Gemini (default gateway/google-vertex:gemini-2.5-flash, switchable to 2.5 Pro). It uses a hardened system prompt focused strictly on Benβs professional background plus a small knowledge base file, and Google safety settings. Guards enforce input validation (length, injection, repetition) and sanitize outputs. Sessions are tracked per session_id with history trimming based on configured turn limits.