Open role
AI & Software Engineer – Aviation Aftermarket (6-Month Internship)
Six month internship building AI that reconstructs an aircraft's maintenance history from tens of thousands of scanned records. Teardown, pre-purchase inspection, lease return.
We're teaching AI to read the paper life of aircraft.
Every aircraft drags a paper trail behind it.
Form 1s. 8130-3s. Work packs. AD and SB status. LLP back-to-birth traces.
Tens of thousands of scanned pages. Zero structure.
The entire aviation aftermarket runs on that paper. Sell a jet, lease it, return it, tear it down: someone still reads every page by hand. It takes weeks.
An engine with a clean back-to-birth trace is worth millions. The same engine without one is worth scrap.
We automate the difference.
The scale of the problem
Tens of thousands
of pages in a single aircraft dossier
Thousands
of components to trace, one by one
Weeks
of manual review, on every transaction
And the findings that get missed are the ones that cost the most.
Where it ships
- Teardown & part-out. Trace every rotable and LLP, or it doesn't sell.
- Pre-purchase inspection. Find the discrepancies before the buyer's engineer does.
- Lease return. Prove the aircraft actually meets its return conditions.
- Asset trading. Engines, APUs, landing gear, LLPs.
- CAMO & MRO. Records versus reality.
Business aviation is the beachhead. Commercial, regional and rotary are next.
The hard parts
Not document search. These are the actual problems.
Read the unreadable
30 year old carbon copies. Inspector stamps. Handwriting. Four languages. No consistent layout, ever.
Resolve identity
One component. Six part number formats. Three serial aliases. A new identity after overhaul, and a different position on a different aircraft.
Build the graph
Components, positions, events, documents and time, modelled so an aircraft's life can actually be queried.
Derive the truth
TSN, CSN and remaining life computed from incomplete evidence, running on two clocks at once: hours and cycles versus calendar.
Never guess
Every output traces back to a page of a document. A hallucinated compliance status is a safety issue, not a bug.
Stack
JavaScript Docker LLMs OCR Computer Vision Knowledge Graphs Vector DBs
Bonus points: NLP PyTorch Graph DBs CAMO / Part-145 / MRO
You
- You like hard problems more than clean tickets
- You build from zero and ship
- You write code other people can read
- You learn fast and ask better questions than you're asked
You don't need an aviation background. You do need to want one.
The hard part of this work is the domain, not the model. The engineers who move fastest here are the ones who already find aircraft interesting, and who read a maintenance manual the way they'd read an unfamiliar codebase. An engineering school background is the natural fit. The rest you learn on real dossiers, next to real engineers.
What you get
Founder, not middle management. You talk to the people making the calls.
A 24 hour feedback loop. Licensed aircraft engineers tell you if you got it right.
Real aircraft. Real deals. Real money on the line.
Apply
Toulouse, France.
Send your CV, your GitHub if you have one, and tell us about the biggest problem you've solved so far in your life.
About Sparengine
Every aircraft has a hidden story. We make it legible.
Sparengine is aircraft intelligence. Decades of scattered records, scans, inspection photos and systems become one connected view of an aircraft: its condition, its history, what the evidence supports and what is missing. Lessors, MROs and traders use it to decide what an aircraft is really worth.
A small team in Toulouse and remote. Real problems, real customers, and ownership from day one.