The posting
At CoMind, we are developing a non-invasive neuromonitoring technology that will result in a new era of clinical brain monitoring. In joining us, you will be helping to create cutting-edge technologies that will improve how we diagnose and treat brain disorders, ultimately improving and saving the lives of patients across the world.
The Role: CoMind One measures the brain from outside the skull. Fifteen scientists (optical physicists, signal processing specialists, physiologists) turn raw optical interference signals into continuous measurements of cerebral physiology a clinician can act on. What slows them down isn't the science; it's the engineering around it. The last head-to-head comparison of two novel methods took six weeks, and most of that had nothing to do with the methods. You'd design the architecture that fixes this: how data, compute, pipelines and evaluation fit together, and the standards that hold it in place. None of it exists yet in any deliberate form, so the shape is an open question. When it works, a scientist describes the comparison they want and gets back metrics, plots and a reproducible record from one command, or one prompt.
At CoMind, all team members work at least 4 days per week from our new Kings Cross offices, plus a flexible work-from-home day.
Responsibilities: Problems, not specifications. How they get solved is yours to decide.
- Reprocessing. Change one stage of the pipeline and you currently pay for a full re-run. Re-running a six-month-old analysis with new parameters should be routine and cheap.
- Evaluation. Comparison and simulation harnesses are rebuilt bespoke every time. They should be something a scientist calls, not something a scientist commissions, including scientists who don't write code, and agents running analysis strands on their behalf.
- Compute and data access. Reproducible environments, sensibly-sized cloud resource, and datasets you can find and trust without asking whoever made them.
- Engineering standards. Testing, CI and repository structure across a codebase written largely by scientists. Knowing where to enforce and where to reduce friction instead matters more than the tooling does.
- The research-to-software boundary. We already have a real interface contract with our software team, which is more than most research groups can say. It should become a gate a method passes, not an event that consumes weeks.
- Regulatory evidence as a by-product. IEC 62304 traceability artefacts generated from CI rather than written alongside it.
- Growing the capability. Mentoring junior developers in the team, and shaping this function as it grows.
AI is fundamental to our culture. It's not just a tool, but a core part of how we work, collaborate, and innovate. We expect all team members to embrace AI in their daily work and continuously find new ways to use it effectively.
Skills & Experience:
- Substantial software or platform engineering experience, a good deal of it building tooling for scientists, researchers or quants rather than for end customers
- Deep Python, and pipeline or workflow systems where correctness and reproducibility mattered as much as throughput
- Cloud compute and storage (AWS preferred), containerisation, infrastructure-as-code, cost-aware provisioning
- Tooling you built that people chose to use. Adoption you earned rather than mandated is the strongest single signal for this role
- The ability to sit with a scientist, understand the analysis well enough to abstract it correctly, and push back when the evaluation itself looks wrong
- A record of raising the standard of the engineers around you without formal authority over them
- Comfort defining a function rather than joining one, on a small team with a real deadline: shipping a usable road early and widening it, rather than designing the complete platform first
Nice to have:
- Regulated environment experience (medical devices, diagnostics, pharma, aerospace), particularly IEC 62304 or software as a medical device
- Time-series or signal processing workloads, physiological data, or scientific instrumentation
- Experiment tracking, model registries or MLOps tooling, aimed at research reproducibility rather than production inference
- Behaviour-driven testing or specification-by-example as an interface between research and engineering
- Introducing agent-assisted workflows into a technical team's actual day-to-day practice
- Having been the first platforms hire into a science organisation before
Benefits:
- Company equity plan so all employees share in the success of the company
- Salary-sacrifice pension scheme
- Private medical, dental and vision insurance (medical history disregarded)
- Group life insurance at 4x annual income
- Comprehensive mental health support, including unlimited access to 1:1 sessions with trained professionals
- Unlimited holiday allowance (+ bank holidays) and one week of remote working per quarter
- Lunch voucher (£10) every day for JustEat and free dinner on those days where you need to work later
- Twice weekly deliveries of fresh fruit and an extensive selection of snacks and drinks
- YuLife subscription, allowing you to turn your daily steps and meditation into discounts at a range of stores
Disclaimer - We use Granola, an AI notetaker, throughout our interview process to help capture notes. It's used only for note-taking during the conversation. Transcripts aren't saved to shared drives or stored externally. Let us know if you'd prefer we don't.
