Top Tech Jobs & Startup Jobs in Toronto

Hands-on field engineer who keeps robot data collection running: stand up and maintain stations, triage and resolve hardware/software/data issues on customer and partner sites, bring up new tasks and sites, and feed operational insights back to model, platform, and hardware teams.

Lead development of a high-performance robotics simulation platform and its compiler stack. Design compute infrastructure and data flow for physics simulation and model training, optimize GPU codegen and JIT compilation, improve differentiable programming support, and contribute to open-source ML compiler projects while collaborating with simulation and robotics engineers.

Own and scale end-to-end physical-data collection operations: scheduling, operator utilization, station uptime, throughput and quality. Recruit and develop operators and shift leads, standardize playbooks and SOPs across sites, partner with Quality and Annotation teams, stand up partner-site deployments, and build reporting that tracks collected hours, quality, and cost.

Design, build, and optimize distributed PyTorch training systems for multi-node GPU clusters. Profile and eliminate performance bottlenecks across data pipelines to GPU kernels, implement low-level CUDA/cuDNN/Triton kernels, tune CPU/GPU/memory/network utilization, and develop monitoring and debugging tools for large-scale training runs.

Design and maintain foundational platform frameworks and internal tooling to accelerate AI research. Collaborate with cross-functional engineering and research teams to build modular, reusable systems, improve workflows, ensure high code quality, and implement organization-wide, low-disruption changes.

Design, prototype, and build robotic hardware systems; develop scalable data-collection platforms; improve simulation fidelity via system identification; maintain in-house hardware infrastructure; collaborate with engineering and simulation teams to deliver robust, scalable physical AI systems.

Own and scale the pipeline that converts raw robot demonstrations (trajectory/video) into training-ready datasets. Design ontologies, run and automate annotation operations (internal and vendor), apply vision-language models for labeling and synthesis, track quality metrics, and close the loop from eval failures to targeted data collection to improve the model.

Develop, optimize, and deploy learning-based robotic manipulation and locomotion systems. Build low-latency teleoperation, train policies with reinforcement and imitation learning, diagnose control-pipeline bottlenecks, and minimize sim-to-real gaps through co-optimization of simulation and real robot behavior while collaborating with a cross-functional team.

Design, build, and maintain petabyte-scale batch and streaming data pipelines and core data infrastructure for robotics foundation model training. Own data models, storage, ingestion, transformation, and orchestration. Standardize pipelines across teleoperation and simulation datasets and collaborate with cross-functional team to support Physical AI development.

Lead the product roadmap for Physical AI, translating research into deployable robotics/autonomy products. Define readiness, success metrics, and reliability thresholds. Align research, engineering, and field teams, prioritize field validation over benchmarks, write clear specs and decision logs, and operate in ambiguous, fast-moving environments.