Location: Delft (hybrid)

Type: Full-time

Start: ASAP

We protect citizens, enterprises, and governments from synthetic media fraud. Everything you see and hear online can now be manipulated — our job is to make sure people can trust what they see. As part of our forensics platform team, you’ll work on the technology that makes large-scale detection possible, from research to production.

You’ll join a small, senior team where your work will have immediate impact, and you’ll have ownership over the systems you build.

• Model innovation: Design and train detectors that combine visual inputs with forensic/context signals into a single system.
• Evaluation & rigor: Build an evaluation framework with clear metrics (e.g., accuracy, calibration, performance breakdowns across attack types) and fully reproducible experiments.
• Low-level optimization: Speed up inference using mixed precision, graph/operator fusion, quantization, ONNX/TensorRT-class toolchains, and other modern approaches — meeting strict performance targets.
• Training at scale: Use distributed training (DDP/FSDP), automated checkpointing, job orchestration, robust data loaders for messy media, and thorough artifact logging/versioning.
• Productionization: Set up a model registry, versioned evaluation, containerized inference, CI/CD, safe rollouts/rollbacks, and telemetry — deployed to both cloud and on-prem environments.
• Reliability & operations: Build for simplicity, observability, and speed; take part in a planned, compensated support rotation.
• Engineering productivity: Develop internal tools and infrastructure to streamline workflows for the whole team.

• Strong Software Engineering foundation: Master’s in Computer Science, Artificial Intelligence, or a related field.
• Production experience: 5–8+ years designing, training, and deploying ML/CV systems end-to-end (Python/PyTorch).
• Model lifecycle ownership: Data preparation → training → evaluation → deployment → ongoing optimization.
• Inference optimization: Skilled in profiling and accelerating models with modern toolchains (e.g., ONNX/TensorRT-class or equivalents).
• MLOps & production skills: CI/CD, containerization, versioning, observability, rollback strategies; experience with experiment tracking/orchestration (e.g., MLflow, Weights & Biases, Kubeflow).
• Distributed training: Scale workloads with DDP, FSDP, or similar approaches.
• Feature fusion: Experience combining multiple signal types where it improves metrics.
• Domain awareness: Familiarity with digital forensics, misinformation threats, or synthetic media — and a strong willingness to deepen expertise.
• Flexibility: Comfortable switching between core ML work and adjacent engineering or infrastructure tasks when needed.
• Mindset & delivery: Thrive in a fast-moving environment; proactive problem-solver; ship, measure, simplify.
• Communication: Excellent written and verbal skills; able to explain complex ideas clearly.
• Independence: Deliver quality work on time without constant oversight.
• Language: Fluent in English.

• On-prem packaging: Containerized deployments and orchestration with Kubernetes.
• Synthetic data: Experience creating datasets to stress-test model robustness.
• Responsible AI: Bias detection, fairness evaluation, explainability.
• Global-scale ML: Multi-region or multi-cloud deployments in AWS.
• Technical track record: Strong GitHub profile, open-source contributions, publications, patents, or public talks.
• Leadership: Mentoring and guiding technical direction.
• Public speaking: Conference or meetup presentations.
• Dutch language: Fluency is a plus.



• A reproducible baseline that improves on our current detector with a robust evaluation suite and targeted error analysis.
• A validated prototype that integrates multiple signal types into the model, with clear insights into each component’s contribution.
• An optimized runtime that meets agreed latency, throughput, and cost goals — with regression tests and basic production monitoring in place.
  

• Mission + ownership: Ship detectors from research to prod and see your work stop real attacks—end-to-end responsibility.
• Company participation: Meaningful equity/virtual shares so you benefit from the upside you create.
• Flexible work: Hybrid (Delft) with flexible hours, maker-time protection, and lean meetings.
• Serious tooling: High-spec workstation plus reserved cloud budget to train/benchmark without friction.
• Research time: Quarterly research sprints and space to prototype ideas that make it into the roadmap.
• Publish & present: Support to speak at meetups/conferences and (where feasible) publish/patent with attribution.
• Performance craft: Time to optimize inference (e.g., quantization/ONNX/TensorRT-class pipelines) without cutting corners.
• Pragmatic on-call: Planned, compensated rotation with clear SLOs, playbooks, and escalation paths.
• Growth path: IC track to Staff/Principal; opportunities to mentor and shape standards.
• Learning budget: Annual budget for courses/books + two relevant conferences per year.
• Home office: Modest stipend for an ergonomic setup; commuting support (public transport or mileage).
• Relocation + visa: Visa sponsorship and relocation support for internationals.

Join us and be part of a company committed to creating a more secure and trustworthy digital future. Apply today to become part of our mission-driven team!