Products That Work
Open source tools. Real problems. Verifiable results.
Adjoint builds specialist software for AI infrastructure, medical physics, industrial quality control, and education. Every product has a repo. Every claim has evidence.
Open source tools. Real problems. Verifiable results.
Adjoint builds specialist software for AI infrastructure, medical physics, industrial quality control, and education. Every product has a repo. Every claim has evidence.
Four domains, one principle: ship tools that solve real problems.
Production observability for AI agents. Logs, traces, and metrics through the LGTM stack (Loki, Grafana, Tempo, Mimir). Framework-agnostic, self-hosted, no vendor lock-in.
A composable library of LLM-native skills for agent workflows. Token-efficient, context-aware, observable. Turns Claude Code, GPT, and other agents into domain specialists.
Workload identity for AI coding agents. Mints signed, cryptographically-scoped bearer tokens (macaroons) instead of handing agents raw credentials — narrowed to one secret, one branch, one agent, one TTL. No get command; plaintext never touches stdout.
Radiochromic film dosimetry. Multichannel calibration, dose analysis, and gamma evaluation for Gafchromic EBT film. Used in proton and photon therapy QA.
Automated tissue quantification for histology and medical image analysis. State-of-the-art AI segmentation with structured CLI output for reproducible measurements.
Monte Carlo radiation transport as a service. Built on EGSnrc, the gold-standard toolkit for radiation dosimetry simulation. Accessible via API for research and clinical use.
LLM-native PLC development toolkit for IEC 61131-3. AI-assisted Structured Text, Ladder Logic, and Function Block Diagram development with version control and CI/CD built in.
Statistical process control for manufacturing and laboratory environments. Control charts, capability analysis, and automated out-of-control detection. Quality monitoring without the enterprise price tag.
Smart building and home automation with a digital twin approach. Calendar-aware, event-driven, fully integrated with lighting, cameras, and voice control.
Extract questions from past exam papers and generate new exam-style questions using AI. Supports UK exam boards (AQA, Edexcel, OCR, WJEC) for GCSE and A-level subjects.
We lead with products, not promises. Consulting follows from demonstrated capability.
Every service we offer is backed by a working tool. If we claim expertise, there's a repo to prove it.
Our core tools are open source. You can inspect the code, run it yourself, and contribute improvements.
MPhys, MSc Medical Physics, MSc Computational Science. Not generalists playing at specialisms.
We contribute back to the tools we use. Active contributor to EGSnrc (NRC Monte Carlo toolkit) and other open source projects.
A sample of what this looks like in practice.
Acted as technical authority on the development of AI training curriculum for NCFE, a national awarding body — shaping how AI is taught to learners across the UK.
Built a RAG-based AI knowledge base grounded in the client's own technical documentation, dramatically cutting troubleshooting time for field engineers.
Converted wiring diagrams and technical documentation into LLM-ready formats, built and open-sourced PLC Toolkit for AI-assisted PLC development, and deployed on-site edge gateways running Kubernetes-based data pipelines.
Open source products. Domain expertise. Real results.
