Available for selected engagements

Hard HMI problems, from schematic to working demo

Biosignal acquisition, haptics, and embedded systems for neurotech teams. I design high-tech proof-of-concepts, rescue fragile prototypes, resolve the hardware-software integration problems that sit between two teams, and act as fractional CTO for companies that need technical leadership before they can hire one full time.

The whole stack is one person: PCB, firmware, signal processing, applications, and mechanics. No integration overhead, no handoff between four specialists who each own a quarter of the problem, and nobody to point at when the layers disagree.

How teams bring me in

Scoped up front, with a written deliverable and a date attached.

POC design

You need a demonstrator investors or reviewers will believe - in weeks, not quarters.

Full stack from blank page to working device: front-end electronics, firmware, acquisition pipeline, and whatever interface makes the result legible to someone who is not an engineer.

6-12 weeks · full stack

System rescue

A prototype that works on your bench and fails everywhere else.

Audit first - noise floor, grounding, timing, power, mechanical tolerance - then a written account of what is actually wrong and a fix for the parts that matter. Often the problem is not where the team assumed it was.

2-6 weeks · audit then fix

Hardware-software integration

Your firmware people and your data people disagree about whose problem it is.

I work on both sides of that boundary, so the question gets answered instead of escalated. Timing, synchronisation, data models, and the interfaces that keep the two halves honest.

ongoing · retainer

Fractional CTO

You need technical leadership before you can justify a full-time CTO.

Architecture and technology decisions, hiring and reviewing engineers, supplier and manufacturing choices, and technical due diligence for investors - while staying close enough to the hardware to catch problems while they are still cheap.

ongoing · 1-2 days per week

Why the timelines are short

Because this is familiar territory. The problems that usually consume the first months of a biosignal or haptic project are ones I have met before, so they get diagnosed rather than discovered.

That groundwork is paid for elsewhere. I maintain CleverHand, an open research platform for biosignal acquisition and feedback - which is where the fundamentals get worked out, on my own time rather than on a client's.

Your design, your data, and your IP stay entirely yours. What carries between projects is method and judgement, never your work.

Known failure modes

Biosignal front ends go wrong in a well-defined set of ways - grounding, common-mode rejection, motion artefact, power coupling. Recognising which one you are looking at is usually the difference between a week and a month.

Architecture settled early

The expensive mistakes in embedded systems are architectural, and they surface late. Timing, data model and power budget get decided deliberately at the start, while changing them is still cheap.

Measured, not assumed

Noise floor, bandwidth and stability get characterised and written down. You get numbers and their limits, not adjectives.

Selected engagements

Neurotech startups, university research groups, and technical leadership roles in London and Cambridge.

NB

Neubond

London2024-Present

Firmware Engineer · Imperial College spin-off

Surface-EMG sensing and haptic feedback for stroke rehabilitation. Redesigned the wearable to reduce its size and simplify the architecture, alongside firmware and mechanical design.

FirmwareEMGMiniaturisation
LY

Lyeons

Cambridge2024-Present

Mechatronics Engineer

A multisensory stress and anxiety management device taken through its full product lifecycle, from battery management and power electronics to research deployment and commercialisation.

Power electronicsProduct lifecycleMechanics
AX

AxonCtrl

London2024-Present

Lead Electronics Engineer

Intramuscular EMG systems for advanced human-machine interfaces. Built 8- and 32-channel amplifier boards, the acquisition GUI, and saline-solution test setups, delivered as an investor-ready demonstrator.

Analog designEMGInstrumentation
UCL

University College London

London2024-Present

Consultant · with Dr Lorenzo Fabrizi

A neonatal haptic research device for studying sensory-motor development in preterm babies. Actuator fabrication, 3D-printed molds, trigger signal processing and EMI testing to clinical-grade reliability.

ActuatorsSignal processingEMI testing
ICL

Imperial College London

London2024-Present

Consultant & Postdoc Researcher · Human Robotics Lab

Z-Block, a high-voltage generator driving soft electro-hydraulic actuators - a niche, high-voltage design problem - alongside multisensory sensing and feedback interfaces.

High voltageSoft actuatorsResearch instrumentation
CN

Cambridge Neuroworks

Cambridge2026-Present

Mechanical Design

An on-ear wearable combining EEG, heart rate, eye movement and motion tracking, aimed at turning balance and falls risk into something clinicians can measure continuously rather than once a year.

Mechanical designWearablesMulti-sensor integration

Some engagements are covered by NDA and are not listed. Happy to discuss scope and references directly.

PhD Bioengineering
Imperial College London
Nature Scientific Data
First-author dataset publication
Marie Skłodowska-Curie
EU research fellowship
EPSRC
Funded grant proposal
ICRA · IEEE WHC
Peer-reviewed robotics & haptics

Got a hard HMI problem?

A few details up front means my first reply can be useful rather than a list of questions. Only name, email, and a description are required - skip anything you are not sure about yet.

About you
About the project
Budget and timeline

Sending opens your email client with the answers filled in - nothing is submitted to a third party. If that does not work, use Copy as text and send it to long.devillard@gmail.com.

Based in London, working with teams across the UK and Europe. Typical reply within 24-48 hours. I take one main engagement at a time, so if the timing does not work I will say so rather than stretch.