Computer vision engineer — leather and defects
Hide scanning, defect segmentation and nesting-aware perception at 30–60 FPS on edge hardware.
OPENWrite code that moves a cutting table
Footeon is a small team building perception and control for physical factories. If you want your model output to become a blade path, a press recipe or a bond window, this is the job.
What working here is like
Four things you should know before applying.
You will go to factories
Nobody builds good manufacturing autonomy from a laptop.
- Regular plant visits in Asia, Europe and Latin America
- Time on the cutting table and the lasting line
- Direct contact with operators and quality engineers
Physical consequences
A bad write is scrap, not a stack trace.
- Shadow mode and evaluation gates are non-negotiable
- Fail-closed design is the default
- Every change is reviewable and reversible
Small team, wide scope
Engineers own a step of the shoe end to end.
- From sensor to setpoint
- From model to edge deployment
- From pilot to customer outcome
Honest engineering culture
We publish methods with results, internally and externally.
- Baselines before claims
- Aspirational marked aspirational
- Post-mortems without blame
Engineering
Roles are listed as [PLACEHOLDER] and will open formally on funding close.
Control engineer — moulding and bonding
Closed-loop control of press recipes, activation temperature and pressure against material-lot drift.
OPENEdge runtime engineer
Rust or C++ runtime on Jetson-class hardware: scheduling, tool sandboxing, OTA and rollback.
OPENSimulation engineer — shoe and line twin
Omniverse-based simulation of cut, mould, last and bond, validated against production outcomes.
PLANNEDFootwear process engineer
The domain anchor: construction specs, adhesive systems, lasting practice and quality agreements.
OPENCustomer solutions engineer
Own deployments from baseline to graduated autonomy inside partner plants.
PLANNEDThis is the product, in one picture
Each node is owned by an agent, and each agent is owned by a small team.
View as table
| # | Node | Agent | Depends on | Status |
|---|---|---|---|---|
| 01 | Twin simulate | Yield-and-Energy | — | SUCCEEDED |
| 02 | Cut and nest | Cut-and-Nest | Twin simulate | SUCCEEDED |
| 03 | Stitch upper | Stitch-and-Upper | Cut and nest | SUCCEEDED |
| 04 | Mould sole | Mold-and-Sole | Stitch upper | SUCCEEDED |
| 05 | Last upper | Last-and-Bond | Mould sole | SUCCEEDED |
| 06 | Bond sole | Last-and-Bond | Last upper | APPROVAL |
| 07 | Inspect pair | Quality-and-Fit | Bond sole | SUCCEEDED |
| 08 | Finish and pack | Robot-and-Assembly | Inspect pair | SUCCEEDED |
Our process
Four steps, roughly two weeks, no take-home longer than three hours.
-
Intro call
Thirty minutes on what you have built and what you want to build next.
30 min
-
Technical deep dive
A real problem from our domain — nesting, control or edge runtime — discussed, not quizzed.
90 min
-
Practical
A scoped exercise on real-shaped data, or a walkthrough of your own prior work if you prefer.
≤ 3 hrs
-
Team and offer
Meet the people you would work with, then a decision within three business days.
3 days
What we offer
Stated as ranges rather than vague promises.
Package
- Competitive salary benchmarked to your market [PLACEHOLDER pending funding close]
- Meaningful early equity with a ten-year exercise window
- Health cover for you and dependants
- Hardware budget including edge devices for home testing
Practicalities
- Hybrid with regular plant travel
- Travel time counted as work time, not as perk
- Four weeks paid leave plus public holidays
- Learning budget for materials science and control theory
Where we are today
Small, technical and close to the factory floor.
- 6 open or planned roles [PLACEHOLDER]
- 3 footwear categories under active design partnership
- 4 countries with partner plants in scope
- 100% engineers who visit factories
Team size, funding stage and incorporation details are [PLACEHOLDER] pending formal announcement.
Concrete problems on the board right now
Not roadmap poetry — the actual open problems.
Nesting under real defect maps
Solve high-quality nests in seconds against scanned hides with irregular boundaries and grain constraints.
Bond-window prediction
Physics-informed models combining humidity, adhesive age, thermal curves and downstream delamination outcomes.
Lasting fit metrology
Sub-millimetre 3D measurement of lasted uppers on a moving conveyor.
Edge scheduling
Deterministic scheduling of a dozen models on one edge unit inside machine cycle time.
Synthetic defect generation
Generate hide scars, stitch skips, adhesive gaps and warped soles that transfer to real detection.
Evaluation for control
Golden datasets and regression gates for models whose failure mode is scrap.
The people you would build for
Cutting rooms, quality offices and plant floors.
“The nesting agent found yield our best pattern cutter could not — and then explained the layout hide by hide. We stopped arguing about scrap and started managing it.”
Leather yield 79.1% → 87.4%
“Bonding is where our warranty costs live. Having an agent watch humidity, primer flash-off and press pressure on every pair — and stop for a human when it wants to move a validated parameter — is the first thing that has actually moved delamination.”
Delamination returns −71%
“We run 40 model changeovers a month. Simulating the nest, the mould recipe and the bond window in the twin before the line starts took hours out of every launch.”
Changeover time −46%
Design-partner scenarios are illustrative and modelled on public footwear-manufacturing benchmarks — named references are [PLACEHOLDER] pending customer approval.
Candidate questions
-
Athletic and sports footwear, casual and leather footwear, and safety footwear, across both brand-owned plants and contract manufacturers. Cut, stitch, mould, last and bond are shared mechanics; the models are tuned per construction — cemented, injected, vulcanised or direct-attach.
-
Agents are bounded by guardrails and schema validation, they cite the datasheet or spec clause behind a decision, and they fail closed to the last known-good recipe. Continuous evaluation against golden datasets gates every model and prompt change in CI before it reaches a plant.
-
Only if you choose. Inference runs on factory-edge servers, and Enterprise deployments can run fully on-premises or in your VPC with no telemetry egress. Brand IP — patterns, lasts, construction specs — is tenant-isolated and never used to train models for another customer.
-
It stops and asks. Every agent runs at a configured autonomy level: observe, recommend, act-with-approval, or act. Bonding, lasting and any parameter under a brand or safety validation defaults to act-with-approval, and every approval is written to an immutable audit log with the sensor evidence that triggered it.
Tell us what you would build first
Send a short note about a system you built that touched the physical world. That is the whole application.