About Cerrion
At Cerrion, we’re building the future of AI-enabled manufacturing. Headquartered in Zurich and backed by Y Combinator and top Silicon Valley investors, we’re building a platform to deploy video AI Agents that can see what’s happening on the production floor, automatically spot problems, and instantly intervene to keep lines running smoothly, at scale. Our technology is live in factories across 15 countries, helping leading manufacturers in glass, beverage, food, CPG, building materials and wood cut downtime by up to 50%. We’ve reached an exciting inflection point — and we’re accelerating fast.
The Role
We’re hiring an Applied Computer Vision Researcher to set and execute the research agenda for Cerrion’s video‑AI platform and turn hypotheses into measurable gains in accuracy, robustness, and latency. We’re looking for a superstar who brings founder‑style ownership across the end‑to‑end research lifecycle — problem framing, data curation, rapid prototyping, experiment design, and rigorous evaluation — balancing scientific depth with pragmatism and a strong bias for shipping. You’ll work directly with our founders and the research team to prioritise the research roadmap, run a fast and reproducible experiment pipeline, and partner with Engineering to turn promising ideas into production‑ready models.
Your responsibilities
Run Fast, Reproducible Experiments: Design and run experiment searches while keeping configs, seeds, and artifacts tidy so that any run can be reproduced with a single click.
Diagnose Bad Runs: Trace failures across data, augmentations, labels, and training dynamics, implement quick fixes, and verify them with targeted evaluations.
Curate & Improve Data: Define datasets, create focused error sets, manage sampling and class imbalance, and partner on labeling and QA to close the loop.
Own Evaluation: Build and maintain a rigorous evaluation suite covering accuracy, robustness, latency, and cost, along with dashboards for trend tracking and failure clustering.
Prototype to Production: Partner with backend engineers to convert promising ideas into production‑ready models, including tests, packaging, canary releases, and monitoring.
Optimise for Cost & Latency: Profile training and inference, choose efficient architectures, and keep GPU hours and serving budgets in check.
Communicate Clearly: Keep the experimental environment clean and well‑documented so that experimental knowledge is preserved and easy to share across the team.
Build the Function: Establish research hygiene — coding standards, reviews, and documentation — and help shape hiring as the team grows.
Desired Skills and Experience
Applied ML & Computer Vision Experience: At least 2‑5 years of hands‑on experience working with real‑world image or video models, with a track record of delivering high‑quality solutions.
Educational Background: Master’s or PhD in Computer Science or a related technical field with a strong foundation in computer vision and machine learning.
Proficient Programming Skills: Exceptional programming abilities in Python and PyTorch, comfortable writing clean training and evaluation loops.
Experiment Operations: Hands‑on experience with modern ML experiment platforms such as ClearML, Weights & Biases, or equivalents.
Data‑Centric Mindset: Practical know‑how in augmentations, class imbalance, sampling strategies, label quality, and active‑learning basics.
Evaluation Rigor: Fluency with metrics such as mAP, F1, IoU, and ROC‑PR, plus calibration, robustness testing, and dataset and version management for reproducible comparisons.
Software Engineering Best Practices: Familiarity with version control (Git), testing frameworks, code reviews, and CI/CD pipelines.
Communication & Collaboration: Strong English communication skills and the ability to turn findings into clear recommendations while working closely with Engineering, Product, and GTM.
Preferred Qualifications
Video & Tracking: Experience with temporal models such as 2D/3D CNNs and transformers, multi‑object tracking, and motion and occlusion handling.
Foundation & Multimodal Models: Familiarity with CLIP‑style pretraining, VLM adapters and LoRA, or lightweight finetuning of foundation models.
Scale & Search: Experience with distributed training (DDP/FSDP), HPO frameworks such as Optuna or Ray Tune, and thoughtful compute budgeting.
Model Optimisation: Experience with distillation, pruning, and quantisation, plus deployment with ONNX Runtime or TensorRT; edge and real‑time constraints a plus.
Data Tooling: Experience with labeling operations, weak supervision, and synthetic data generation and augmentation pipelines.
Domain Context: Prior work in video analytics for industrial or safety‑critical environments.
Kaggle Competition Participation: Previous involvement in Kaggle competitions, demonstrating practical experience in tackling real‑world machine learning problems.
Why Cerrion
Ownership & impact: Work on hard, real‑world vision problems that directly improve manufacturing for some of the world’s largest producers. Your models will run in factories across 15 countries.
Growth path: This role is designed to evolve as our research capabilities grow — you’ll shape the technical direction and help build the team.
Culture: We operate with flat hierarchies and a get‑things‑done mentality. You’ll work with ambitious colleagues from all over the world in a collaborative environment where the best idea wins.
Compensation & equity: Competitive salary, meaningful stock option package and comprehensive benefits.
Location: Modern office in the heart of Zurich; enjoy the quality of life of Switzerland and the dynamism of a high‑growth startup.
If you’re passionate about pushing the research frontier in applied computer vision, running rigorous experiments at pace, and seeing your work deployed at scale in real factories, we’d love to meet you.