Website:
zapdoslabs.com
Job details:
ABOUT ZAPDOS LABS
Zapdos Labs builds AI video agents for safer and more efficient industrial operations. Our systems connect to existing cameras to detect safety risks, monitor operational sequences, analyze worker and vehicle movement, and understand events across multiple cameras.
We combine modern multimodal models with precise computer vision so that every alert is accurate, relevant, and supported by video evidence.
THE ROLE
We are looking for a Senior Computer Vision Engineer to own core parts of our perception stack, from raw video and spatial calibration to tracking, event understanding, evaluation, and production deployment.
This is a hands-on role. You will build models and pipelines, design experiments, investigate failures in real footage, and turn research into systems that operate reliably on factory floors.
WHAT YOU WILL DO
• Design and productionize detection, segmentation, pose, tracking, re-identification, and activity-recognition systems
• Build reliable pipelines for continuous video from fixed cameras in challenging industrial environments
• Develop multi-camera tracking and event association across viewpoints, zones, and time
• Use camera calibration, scene geometry, trajectories, and 3D representations to reason about physical interactions
• Detect events such as worker-vehicle proximity, PPE non-compliance, restricted-zone entry, unsafe actions, and procedural deviations
• Combine learned models, vision-language models, temporal reasoning, geometry, and deterministic filters
• Translate customer SOPs and safety requirements into measurable perception tasks
• Define evaluation datasets and metrics for event accuracy, false-alert rate, latency, robustness, and operational usefulness
• Build testing and simulation workflows using digital twins before deployment to customer cameras
• Improve performance across changing lighting, occlusion, camera quality, site layouts, uniforms, equipment, and behavior
• Optimize inference for edge or cloud deployment under real-time resource constraints
• Analyze production failures and create systematic improvements to models, data, and evaluation
• Guide annotation strategy, active learning, hard-example mining, and dataset quality
• Mentor engineers and help establish the technical direction of the computer vision platform
WHAT WE ARE LOOKING FOR
• Deep practical experience building computer vision systems that operate on real images or video
• Strong knowledge of object detection, segmentation, tracking, temporal modeling, and model evaluation
• Proficiency with Python and modern deep-learning frameworks
• Experience taking models from experimentation through production deployment
• Strong understanding of precision, recall, calibration, thresholding, and error analysis
• Ability to diagnose failures caused by data quality, domain shift, geometry, infrastructure, or system design
• Experience working with large video datasets and annotation pipelines
• Strong software engineering fundamentals, including testing, profiling, versioning, and reproducibility
• Ability to communicate model limitations and tradeoffs clearly to product and customer teams
• Comfort making technical decisions in a fast-moving early-stage company
• Careful handling of confidential customer video and operational information
NICE TO HAVE
• Experience with multi-object or multi-camera tracking and person or vehicle re-identification
• Experience with camera calibration, homography, 3D reconstruction, digital twins, or synthetic data
• Familiarity with vision-language models, video-language models, or agentic multimodal systems
• Experience with action recognition, sequence compliance, anomaly detection, or trajectory analysis
• Experience optimizing models with CUDA, TensorRT, ONNX, DeepStream, or similar technologies
• Familiarity with RTSP cameras, GStreamer, video codecs, and live-stream ingestion
• Experience deploying inference workloads to GPUs or edge devices
• Background in robotics, autonomous systems, manufacturing, logistics, construction, or industrial safety
• Research publications or meaningful open-source work in computer vision
• Experience leading technical projects or mentoring other engineers
WHAT SUCCESS LOOKS LIKE
During your first months, you will:
• Understand the complete path from camera footage to an operator-facing alert
• Establish clear baseline metrics for one or more production detection tasks
• Identify the most important sources of false positives and missed events
• Ship improvements that can be verified on representative customer footage
• Strengthen evaluation, deployment, and monitoring practices
• Help the team support new sites and camera configurations without rebuilding every pipeline manually
COMPENSATION AND STAGE
Zapdos Labs is an early-stage company building production AI systems for industrial customers. Compensation includes cash based on experience and commitment, meaningful equity reflecting senior technical ownership, flexible remote work, and direct influence over research, architecture, and product.
WHY THIS ROLE
Industrial computer vision has different constraints from benchmark research. Cameras are imperfect, environments change, events are rare, and false alerts quickly destroy user trust.
You will work on these problems directly and see your systems deployed in operational environments. Your work will influence whether an unsafe interaction is recognized, whether a procedural error is caught, and whether an operator receives useful evidence in time to act.
Click on Apply to know more.
