Joel Beedle

Joel Beedle

Agentic systems engineer, UK.

I lead the technical build of AI products. Most recently, a whole new product line at Procure Ai, from architecture to product decisions.

— agents · Gathering · tap to scatter

See the workEmail me

About

Tap a word to see what it attends to.

Use the left and right arrow keys to move between words, Home and End to jump to the first or last, and Escape to clear.

I’m Joel, an agentic systems engineer at Procure Ai in the UK. I led the technical build of a whole new product line there, making the product calls as well as the architectural ones. Before that I was a software engineer at Bendi, built a triage AI for a real-patient study, and presented a swarm simulator at ACSOS in Denmark.

Waiting for a word.

Attention from SmolLM2-360M, layers 13–24, all heads averaged; the start-of-text token removed. Computed from this paragraph ahead of the site build.

A map of my work.

Every point is a word from these write-ups, placed by meaning. As each one scrolls up, the map turns to show where it sits.

Work

Procure Ai · Agentic systems engineer · Nov 2025 — now

A new agentic product line

Procure Ai is building the agentic operating system for procurement: more than fifty autonomous and collaborative agents that take on sourcing, purchasing and supplier work for enterprise teams.

I led the technical implementation of a whole new product line, from the first architecture through to production, and made key product decisions along the way. I worked closely with Customer Success, Sales, Operations and the wider engineering team to shape it.

Day to day I engineer the orchestration and tooling behind the platform's autonomous agents: the full-stack infrastructure that lets AI agents run reliably in production.

Bendi · Software engineer · Nov 2024 — Nov 2025

Multi-agent AI for supply chains

Bendi is an AI-powered supply chain platform that helps companies see and manage environmental and social risk deep in their supplier networks.

I architected multi-agent AI systems, built document processing pipelines that chunk complex corporate questionnaires for retrieval-augmented generation, and built AI-enabled n-tier mapping for large, multi-level supplier datasets.

I also built and tested event-driven systems across GCP and AWS, worked on the cloud infrastructure and deployments, and led a UX overhaul of the platform for enterprise users navigating complex supplier networks.

  • GCP
  • AWS
  • Pub/Sub
  • Cloud Run
  • Terraform

Institute for Safe Autonomy · Research software engineer · Jul — Nov 2024

Diagnostic AI for A&E triage

DAISY is a diagnostic AI system for emergency department triage, used in a study with real patients at York & Scarborough Teaching Hospitals.

I led its implementation and deployment on an Android tablet for patients: I designed the architecture and the security around the deployment, built a REST API on the tablet with a static Next.js frontend served as a web view, and built the Java Spring Boot backend.

I also shaped the interface to make it friendlier for patients, following NHS design principles.

  • Next.js
  • React
  • Java
  • Spring Boot
  • Android
About the project

University of York · Research assistant · 2024

A swarm algorithm simulator

SALSA is an aerial swarm algorithm simulation platform I created for my Masters dissertation: write a swarm algorithm, then watch it run in a 2D world.

A funded research internship let me turn it into a paper, which I presented at the 5th IEEE International Conference on Autonomic Computing and Self-Organizing Systems (ACSOS 2024) at Aarhus University in Denmark.

It is written in C++ with OpenGL and compiles to WebAssembly, so the whole simulator runs in the browser.

  • C++
  • OpenGL
  • WebAssembly
Run the simulatorSource

Say hello.

Open to interesting problems, collaborations, and good coffee.

beedlejoel@gmail.com
GitHubGitLabLinkedIn

Based in the UK · © 2026 Joel Beedle

MEng Computer Science, University of York · First Class Honours, 2024