UPLABS is a research lab and venture studio. We take problems most consultancies refuse — physical AI, novel hardware, regulated frontiers — and turn them into shipping companies. Six co-founded in four years. Every one still operating.
A research squad embedded with your team for 6+ months. We tackle the hard, multi-year problems your roadmap can't fit.
// uplabs.config.ts export const rd = { squad: 5, cadence: "bi-weekly", outputs: ["paper", "patent"], equity: false, // optional };read the spec
Fourteen weeks. One question. A working prototype, a technical memo, and a board-ready go/no-go recommendation.
# sprint.toml [phase] discovery = "weeks 1-3" build = "weeks 4-10" evaluate = "weeks 11-14" gate = "go/no-go"read the spec
We co-found one company per year with operators we believe in. Capital, founding team, and the lab's full research stack.
// term-sheet.json { "check": "€250k–€2m", "equity": "15–25%", "role": "co-founder", "runway": "18 mo" }read the spec
Closed-loop hydroponics that learn crop-by-crop. 90% water reduction in arid climates.
Untethered AUVs for offshore inspection, repair, and ecosystem restoration.
Open-source AI-compatible smart PDU. The Lego of remote energy & environment.
Smart controllers + ML for industrial kitchens. Predicts failure before the line stops.
A T5 transformer trained on millions of patented reactions. Predicts before bench run.
Acoustic AI that detects FPV drones from passive audio — 97% accuracy.
You send a brief. We send back a fit memo, a named lead engineer, and a fixed-fee scope. No 12-page proposal theatre.
Literature scan, baseline benchmarks, working list of unknowns. We tell you which unknowns are worth the cost of removing.
Squad ships. Bi-weekly demos with the actual code, not slides. Your team sits in if they want — most do.
Working prototype, technical paper, IP packet, and a recommendation. Optional: continue with us, or take it in-house.
A field survey of where the bottleneck has actually moved — from algorithms to electrons, from FLOPs to fab capacity, from VCs to hyperscaler treasuries. Twelve calls and twelve we got wrong.
That's our beat. Send us a one-pager. We'll tell you in 72 hours whether we want to take it on, refer you elsewhere, or talk you out of it entirely.