Applied Scientist I

Applied Scientist I to design and train cutting-edge machine learning solutions related to digital identities. Join our team and work on challenging problems in deepfake detection, bias mitigation, document understanding, anomaly detection and/or efficient ML.

About the Team:

You'll be joining the team leading Entrust's Identity portfolio, formerly known as Onfido (an AI- powered digital identity solution). Our technology helps businesses verify real identities using machine learning, ensuring secure remote customer onboarding. By assessing government- issued identity documents and facial biometrics using state-of-the-art machine learning, we provide companies with the assurance they need to operate securely while allowing people to access services quickly and safely.

Our Applied Scientist team consists of about twenty machine learning scientists. The team is supported by an ML Ops team that provides state-of-the-art tooling (including AWS, Encord, Ray, PyTorch Lightning and Weights & Biases). The Applied Science team works closely with product engineering to deploy models to serve our worldwide customer base.

Position Overview:

We are looking for a Applied Scientist I to design and train cutting-edge machine learning solutions related to digital identities. Join our team and work on challenging problems in deepfake detection, bias mitigation, document understanding, anomaly detection and/or efficient ML.

What you will be doing:

• Push the frontier of research in areas such as deepfake detection, bias mitigation, fraud/anomaly detection, face matching, document understanding, and efficient on-device ML. 
• Publish research results in national and international conferences and scientific journals.
• Work with product and engineering to improve our world-class identity-focused products.

Representative work:

• Implement bias-mitigation strategies to build fair face-matching and deepfake-detection models.
• Train and benchmark large-scale vision-language models for document extraction. Train a multi-modal document understanding model from scratch using synthetic data.
• Optimise LoRA adapter latency in PEFT/Triton.
• Profile, debug and improve model training speed on multiple GPUs.
• Create a large-scale dataset for deepfake detection.
• Experiment with multimodal models to detect fraud.

You may be a good fit if you:

• Have strong experience in machine learning and computer vision.
• Have a strong record of successfully delivering high-performance ML-driven products.
• Have a deep understanding of machine learning theory.
• Have strong coding skills in Python and PyTorch.
• Care about building fair and cutting-edge machine learning products.

Strong candidates may also have:

• Technical experience in one or more of the following areas: face matching, bias mitigation, anomaly detection, document understanding or on-device ML.
• Published at top-level machine learning conferences.
• Experience optimising (distributed) training code.

For US roles, or where applicable:

Entrust is an EEO/AA/Disabled/Veterans Employer

For Canadian roles, or where applicable:

Entrust values diversity and inclusion and we are committed to building a diverse workforce with wide perspectives and innovative ideas. We welcome applications from qualified individuals of all backgrounds, and we strive to provide an accessible experience for candidates of all abilities.

If you require an accommodation, contact accessibility@entrust.com.

Recruiter:

Jack Steib

jack.steib@entrust.com