Computer Vision Data Scientist

About Tasq.ai

Tasq.ai is a data company that fuel AI development. We build the annotation, evaluation and data-filtering infrastructure that physical AI teams use to train, evaluate, deploy and monitor their models, with a strong focus on robotics and collaborative robots .

Our edge comes from speed at scale (parallel, elastic processing), smart filtering and failure clustering that find the data that matters, and statistical engines that turn noisy judgments into reliable signal.

The role

We're hiring a Computer Vision Data Scientist with about 3 years of experience who has taken CV models all the way from raw data to production, and kept them working there.

We're looking for fast and agile problem solver. independent data scientist that can solve a problem end2end. 

What you'll do

• Design, train and ship CV models for video and image understanding: detection, segmentation, tracking, action and temporal segmentation of robot manipulation tasks.
• Own the full lifecycle: data selection and curation, labeling specs, training, offline evaluation, deployment, and production monitoring.
• Build evaluation methods and metrics that hold up in production, including quality scoring of human and model annotations.
• Develop data-filtering and failure-mining tools that find edge cases, cluster model failures and prioritize what to label next.
• Use pre-trained and foundation models (vision-language models, video models) as pre-labelers, judges and feature extractors, and know when to fine-tune.
• Optimize models for cost and latency at scale, working with engineering on serving, pipelines and MLOps. 

What you'll bring

• 3+ years of hands-on experience as a data scientist or ML engineer focused on computer vision.
• At least one CV model you delivered end to end into production, from data to deployment to monitoring, and can walk us through in depth: what broke, how you measured it, and what you changed.
• Strong Python and deep-learning skills in PyTorch (or similar), with modern architectures such as CNNs, vision transformers and detection/segmentation frameworks.
• Solid grasp of evaluation: choosing metrics, building test sets, error analysis, and spotting data leakage and distribution shift.
• Comfort with data at scale: large image and video datasets, labeling workflows and data quality issues.
• B.Sc. or higher in Computer Science, Electrical Engineering, Math, Physics or a related field.
• Clear communication in English, and the ability to explain trade-offs to engineers and customers alike.

Nice to have

• Video understanding experience: temporal models, action recognition, multi-object tracking.
• Robotics or physical AI data: manipulation, egocentric video, multi-camera or 3D/depth data.
• Hands-on work with vision-language or video foundation models, including fine-tuning and prompting them as evaluators.
• Active learning, data curation, or human-in-the-loop labeling systems.
• Statistics for noisy labels: inter-rater agreement, confidence estimation, rater weighting.
• Model optimization for deployment: quantization, TensorRT/ONNX, edge or GPU-efficient inference.
• An M.Sc. or publications in computer vision.

Why join

• Work at the frontier of physical AI, on real robot data that leading teams depend on.
• Real ownership: your models go to production and you see their impact directly.
• A small, senior team where you work closely with founders, R&D and customers.