AI Engineer - Biometrics

Didit (YC W26); Barcelona; On-site; EUR 90,000 - 120,000; Full time

Benefits offered by this job

On-site Barcelona office

Research publication opportunities

Competitive compensation

Job summary

Didit Labs, the in-house AI lab of Didit, builds state-of-the-art biometric models shipped to production for millions of users. We seek an AI Engineer to own face biometrics end to end, from research to a fast, reliable service in production.

You will lead iBeta certification efforts and build defenses against spoofing, deepfakes, and injection attacks, with latency considerations for on-device and server environments.

Qualifications

• Deep computer vision expertise focused on faces: liveness, anti-spoofing and recognition.
• Proven track record of building and shipping face models in production.
• Strong deep learning fundamentals with PyTorch and model training pipelines.
• Hands-on experience taking a PAD or liveness system through iBeta Level 2 or Level 3.
• High-ownership mindset and comfort working in a small, fast-moving team.

Responsibilities

• Own our face biometrics models: liveness detection, anti-spoofing and face recognition.
• Lead the iBeta certification project end to end: test plan, attack-data collection, lab submission and fixes.
• Build detection for injection attacks and deepfakes: virtual cameras, emulators, face swaps and generative video.
• Take research ideas into production, on-device (CoreML, TFLite, ONNX) and server-side, with tight latency.
• Benchmark against NIST FRTE and measure demographic fairness.
• Publish research with our team at venues such as IJCB, CVPR workshops and BIOSIG.

Skills

Face biometrics

Liveness detection

Anti-spoofing

Face recognition

Production deployment

Tools

PyTorch

CoreML

TFLite

ONNX

Job description

At Didit Labs, our in-house AI lab, we build state-of-the-art biometric and fraud models that are shipped to production and used by millions of people every month. We're looking for an AI Engineer to own face biometrics end to end, from research to a fast, reliable service in production.

Didit is often called "the Stripe of identity". We provide the infrastructure for identity verification and fraud prevention: one API to authenticate, verify and monitor users and businesses. Our platform covers KYC, KYB, biometric liveness, AML screening, transaction monitoring and wallet screening.

• The only identity provider formally attested by an EU member state as safer than in-person verification
• 220+ countries, 14,000+ document types and 48+ languages
• 3,000+ companies worldwide, from fintechs and crypto exchanges to marketplaces, healthcare and AI labs
• Backed by Y Combinator and Robinhood Ventures, with offices in Barcelona and San Francisco

The mission

Make Didit's liveness and face matching the hardest in the market to fool, and prove it with independent certification. You'll take our presentation attack detection (PAD) through iBeta ISO/IEC 30107-3 Level 2 and Level 3, and build defences against injection attacks and deepfakes, the next battleground in identity fraud.

What you'll do

• Own our face biometrics models: liveness detection, anti-spoofing and face recognition
• Lead the iBeta certification project end to end: test plan, attack-data collection, lab submission and fixes
• Build detection for injection attacks and deepfakes: virtual cameras, emulators, face swaps and generative video
• Take research ideas into production, on-device (CoreML, TFLite, ONNX) and server-side, with tight latency
• Benchmark against NIST FRTE and measure demographic fairness
• Publish research with our team at venues such as IJCB, CVPR workshops and BIOSIG

What we're looking for

• Deep computer vision expertise focused on faces: liveness, anti-spoofing and recognition
• A proven track record of building and shipping face models in production
• Strong deep learning fundamentals with PyTorch and model training pipelines
• Hands-on experience taking a PAD or liveness system through iBeta Level 2 or Level 3
• A high-ownership mindset and comfort working in a small, fast-moving team

Nice to have

• Published research or competitive ML achievements (e.g. Kaggle)
• Experience with biometric systems running at scale
• Experience with injection or deepfake detection (CEN/TS 18099) or NIST FRTE/FATE submissions

What we offer

• Full ownership of face biometrics at a fast-growing, YC-backed company
• Real attack data at a scale few companies can offer
• Freedom to publish your research
• A seat in Didit Labs, working on-site in our Barcelona office with a small, elite team
• Competitive compensation, discussed individually during the process

If you want to build the hardest-to-fool liveness system in the market, we'd love to hear from you.