Computer Vision + Reinforcement Learning Engineer

Please only apply if you have previous Nvidia Omniverse, Computational Fluid Dynamics, or VFx experience

About OmniPath

OmniPath is building high-performance, real-time perception and reinforcement-learning systems for sports analytics and industrial energy optimization.

On the sports side, we’re deploying a next-generation CV + simulation stack (object tracking, event inference, physics-aware embeddings, 3D overlays) for live broadcast and sportsbook integrations. Curling is our first vertical, but the underlying architecture generalizes to multi-sport tracking, play prediction, and virtual broadcast augmentation.

On the industrial side, we extend the same RL and digital-twin backbone to rotating-asset optimization—compressors, pumps, turbines—using GPU-accelerated physics, real-time inference, and Omniverse-based simulation staging.

We’re Calgary-based, NVIDIA Inception-aligned, and shipping real GPU-accelerated production code across sports and energy environments.

What You’ll Do

You will own one or more vertical slices across real-time CV, RL training loops, and 3D/Omniverse deployment:

Computer Vision

Multi-object tracking: detectors, trackers (ByteTrack/DeepSORT), ID stitching, occlusion recovery

Event + state inference: play/shot classification, trajectory smoothing, impact/terminal state extraction

Camera calibration, homography recovery, field-of-play normalization

Latency-tuned GPU inference (batching, mixed precision, TensorRT optimizations)

Reinforcement Learning

Gymnasium-compatible environments for sports and industrial settings

PPO/SAC baselines for shot/play recommendation, or control policies for rotating equipment

Sim-to-live transfer, reward shaping, evaluation harnesses

TorchScript/ONNX exports integrated into real-time pipelines

3D / Digital Twin / Omniverse

USD scene prep from Blender/Unity assets

Real-time overlays for sports broadcasts

Physics-informed pipelines for industrial digital twins (Omni.PhysX, Flow, Replicator)

Infrastructure & Tooling

Annotated video/debug artifacts, metrics dashboards, validation sets

(Optional) AWS GPU infra: EKS, EC2, MediaConnect/IVS, S3 pipelines

Profiling: Nsight, TensorRT, GPU scheduling

You Might Be a Fit If You Have

Strong Python engineering fundamentals

CV experience: OpenCV, PyTorch, Detectron2/YOLO, object tracking, calibration

RL familiarity: PPO/SAC/IMPALA, Gymnasium/Gym, RLlib, SB3

Data tooling: NumPy, Pandas, Parquet, DVC

(Bonus) CUDA/TensorRT, AWS GPU infra, Unreal/Unity, Blender → USD, Omniverse Composer

Time Commitment

5–10 hours/week to start. Async-friendly, milestone-based.