About

Hi — I’m Philip Redford-Jones (Phil).

I’m a Software Developer and AI Engineer with experience building production machine learning systems, cloud infrastructure, and data-driven products. My work sits at the intersection of applied machine learning and software engineering.

I’m particularly interested in natural language problems, reinforcement learning, games, and graph-based methods. You can see what I’m currently working on on my GitHub.

Bio

I studied Physics and Chemistry at University College London from 2014 to 2017, graduating with a BSc in Natural Sciences. During this time, I researched computational methods for drug discovery, which sparked my interest in computing and machine learning.

After several years strengthening my software engineering skills, I refocused on artificial intelligence and completed an MSc in Machine Learning at UCL. During my master’s, I developed a strong foundation in deep learning, natural language processing and LLMs, reinforcement learning, graphical methods, and multi-agent systems.

I’ve lived and worked in London and France, and I’m now based in Pretoria, South Africa. Professionally, I’ve worked at the London Stock Exchange Group as well as a number of startups and scaleups.

In my current role at Prolific, I work as an AI Engineer building and deploying production-grade machine learning systems that support real-time audience targeting, fraud detection, and intelligent study matching at scale.

My work focuses on deploying LLMs and traditional ML models on Kubernetes for low-latency, high-availability inference. I’ve led the development of real-time fraud detection pipelines and designed embedding-based retrieval systems for searching and recommending participants for studies and AI data-labelling tasks.

I’ve also set up and operationalised Pinecone as a vector database, alongside MLflow for experiment tracking, model versioning, and governance, helping to establish a more mature ML lifecycle across the organisation. A key highlight was reducing the latency of Prolific’s real-time audience finder tool from ~16 seconds to ~2 seconds through model optimisation, caching, and service-level improvements.

Alongside product-focused work, I collaborate closely with the platform team to improve the data platform, deployment patterns, and service architecture, ensuring ML systems are reliable, observable, and scalable in production.

While my core strength is AI product development, I also have experience across the full stack.

The best way to reach me is via email at philiprj2@gmail.com.