Substrate Bio
Member of Technical Staff, Physical AI
How your CV stacks up
Upload your CV to see how well it fits this job role
?%
Member of Technical Staff, Physical AI
Engineering Director, Lab Perception Layer
The opportunity
Substrate is building a laboratory that runs itself. Robotics move the samples while instruments take the measurements. For an autonomous lab to function without human oversight, it must perceive its activities: every camera stream, sensor log, and precision measurement must be captured and analyzed in real time—this perception layer still doesn’t exist.
You would architect it. Your work lets the lab detect errors—the missed dispense, blocked path— the instant they happen, flagging them to the team before they corrupt datasets. This isn’t just error detection; it’s ensuring the provenance of every experiment at source. Without accurate, granular data recording, Substrate’s foundation models wouldn’t have the high-quality, context-rich training data they need.
You’re building a live perception layer before the lab enters semi-automated operation. Anydata lost isn’t recoverable.
About Substrate
Substrate is developing the critical infrastructure layer between AI and biological discovery: an autonomated lab tailored for AI. Unlike cloudbased systems or contract research organizations, Substrate is the intermediate software pipeline that executes hypotheses and records structured, AI-ready data.
Founded by four veterans with backing from investors, we’re constructing our first lab at 20 Triton Street, London. We generate scalable high-quality data with embedded provenance—opening up possibility for foundation models in biology that have faced imprecise or unlinked data before.
The role
Your focus
You’ll design and implement the lab’s self-perception system:
- Hardware onboarding: Integrate every sensor and camera stream into a live, spatially mapped data model that reflects the physical environment.
- Real-time error detection: Deploy computer vision models (VLMs, LLMs, or vision pipelines) that watch lab operations, detecting anomalies as they occur before they propagate.
- Edge deployment: Process and respond locally—real-time error correction cannot rely on cloud pipelines.
Your layer is foundational. All higher-level software (workflow execution, scientific workflow orchestration, and learned intelligence) builds atop your data’s cleanliness, completeness, and structure. You’ll collaborate with:
- Engineers who write the core lab-operations infrastructure code
- Scientists utilizing assay models
- Teams giving the data its context (assays, annotations, and research context)
First 90 days
FIRST 30 DAYS
- Deploy the first operational cameras and sensors onto the semi-autonomous lab operations—no activity happens without a corresponding capture pipeline.
- Finalize a single, timestamped data schema that works for hardware, science, and infrastructure teams.
Reasons to use Rodeo
Start with a chat, not a search bar
Grad scheme, placement, apprenticeship? Not sure what you want yet — that's fine. Your agent talks it through with you and turns "I have no idea" into a shortlist.
It searches the market for you
Every day your agent scans the market matching roles against what actually matters to you, not just keywords on a CV.
Only hits
No noise. No "maybe this fits." Just roles with a clear explanation of why they're right — and where to focus when applying.
DAYS 30 TO 60
- Deploy real-time error detection on a live instrument, alerting the team at first deviation (e.g., misfired pipette, instrument occlusion).
- Expand capture to instrument logs and lab movement data (robot payload transfers, processing timelines).
DAYS 60 TO 90
- Integrate environmental sensors (climate, humidity).
- Optimize and harden the edge pipeline—camera handling, inference latency, fault tolerance.
- Surface pre-structured, captured data to the metadata layer, ready for AI model consumption as automation scales.
Who you are
You thrive in the gap between digital and physical systems, ready to ** ÄndOwn implicit error modes**—calibration clashes, sensor drift, momentary mis-alignment.
You have industrial-strength hardware-aware vision at scale, preferably shipping software with tactile designs:
- Expertise in sensors, hardware integration—adapter cables, synchronization, telemetry workflows, or deployment environments (e.g., lab automation, robots, manufacturing).
- Production machine vision—VL models crafted for adversarial physical conditions (e.g., 3D Machines checkpoints, drone photogrammetry, or manufacturing SLAM constraints).
- Pragmatic mindset—Pinpointing VLM’s anomalies, like global context in the wrong task mode, or requiring repeatable retraining templates.
Requirements
Must have:
- Four or more years building high-reliability perception systems within dynamic physical environments (robotics, ATG, scientific instrument monitoring…).
- Hands-on sensor/computer vision onboarding—telemetry pipelines, camera calibration, time-sync, perspective correction, and hardware synchronized data steam management.
- Fluency in CV production pipelines—dozens of model configurations evaluated for latency/drift.
- Strong FPGA/System on Chip development skills to edge-process data local to events (e.g., real-time event detection, }];
- Preference for programmable compute work (MCU > Nvidia Jetson > bare-metal---
- Comfort managing systems integration at the boundaries between SDK vendors and internally written software.
Nice-to-haves:
- Experience scaling production sensor fusion or SLAM pipelines (localization path optimization under hardware multirobot constraints).
- Direct familiarity with industrial robotics data buses (e.g., bite...)
- Science/assay data handling awareness (collected data shape, research model ID inclusion, human oversight as an operation graceful checkpoint).
- Prior founding team stints—exploiting domain growth signals as development opportunities, advocating and stopping false/ wasteful- parallel tracks
Why this is unusual
Get help with your application
Your very own career expert that helps elevate your application to the next level.
Unlike traditional physical-robotics perception roles focusing on self-navigation models, here you’re ensuring 100% data validity for biological datasets. If the AI foundation model trains on a misfiltered sample edge, the artifacts track forever. Unlike STEM service providers, no engineering or research siloes here.
Your work is **structured as a sufficient capturer of atoms (e.g., robot hand understanding). Your architecture choices happen fast—there are no central points of failure beyond alerting. Solutions must design for graceful scale as the autonomous lab transitions from semi-manual to fully within-capture autonomy.
How we work
Unlike fully distributed teams here three days week are in-lab, interacting directly with the robots, instruments, and sensors. The rest of the team works flexibly in their geographic locations.
We are light on conventional meetings, but:
- Occasional kickoffs on Mondays
- All-hands on Thursdays
- Once-per-day streamed team sync
- Off-sites every quarter to clarify vision artifacts
We are a small team preserving sanity through shared innovation documentation and fast-response integration of the constructive ideas, regardless of seniority.
Embedded benefits in the UK:
- 32 days annual leave + international holidays
- 8% company pension contribution, with primary health insurance benefits with Bupa
- Team coverage grows swiftly when the engineering, science, and data science teams expand
The team you’ll join
You’ll join our four founding engineers, four with a process that replicates between hardware and continuous customer interaction:
- Mostafa ElSayed (Founder and CEO, [Automata بخسبط), responsible for lab architecture.
- Oli Hoy (*Software Operations Leader, bench-level infrastructure + assay timeline guidance.
- Alex Mordukovov (Intelligence/Science Layer), orients the models toward the task
- Anna Huygues-Despointes drives our commercial partnerships and customer success, passing back measurables from real customer feedback.
You will report to a Founding Engineer who owns primary infrastructure. When we cross a tipping point of our lab’s complex automation nuances (50->250 hardware edge instances), the team will already be double its size.
Process
First we screen qualifications and take two-hours to exchange cultural fit over video. Shortly after, participants must run a 1–2 hour live technical session to discuss solutions with the team, reflecting on equally first-hand design/SLAM issue walks.
We base our hiring on mutual intent toward merit and scalability contributions. Starts are open to every background—that talent unlocks diverse perspectives reinforces that reality.
“It took my CV and asked me questions relevant to understanding what kind of jobs to suggest for me. Suggestions were almost perfect. Jobs were exactly what I’ve been looking for.”
Jessica, London
Skills
Location