Active Initiatives & Research Sprints

Stop theorizing.

We are publicly defining Canada’s robotics and embodied intelligence landscape. Each sprint is live, structured, and moving forward. Miss the window, and others will set the reference without your input.

The Sprint List

Choose how you collaborate with CPAI

Claim the sprint that best matches your skills. Each initiative has clear deliverables, dedicated leadership support, and a tight handoff rhythm.

Join the project
Offline Sensor Health Analysis

SensorGuardian is an open-source, offline tool for analyzing ROS 2 rosbags and generating deterministic sensor health reports. It helps teams detect silent sensor degradation, synchronization issues, and data integrity problems before deployment or debugging.

  • The Task: Build deterministic health checks for cameras, LiDAR, IMU, and GPS; parse rosbag2 data to ship human-readable + machine-readable reports; validate checks against real-world failure cases; maintain golden test bags and regression tests.
  • Ideal for: Robotics / ROS 2 engineers, QA & validation engineers, autonomy engineers focused on reliability and safety.
  • Status: Active development with architecture and v0.1 scope defined.
Join the project
Campus outreach cohort

A structured collaboration with university students who represent CPAI on campus, distribute opportunities, and recruit contributors for projects and research sprints.

  • The Task: Distribute CPAI opportunities through campus channels; post about CPAI activity monthly; coordinate with at least one student group; relay feedback and interest signals to the CPAI team.
  • Ideal for: University students in Mechanical, Electrical, Mechatronics, Computer Science, or closely related disciplines with access to student communication channels.
  • Status: Applications open for the current cohort.
Join the project
Knowledge transfer series

An invitation to experienced engineers and technical leaders to share real deployment constraints, failure modes, and hard-earned tradeoffs with students and early-career builders.

  • The Task: Share a focused talk, case study, or lessons learned on real systems, deployment, testing, or scaling.
  • Ideal for: Practicing engineers, technical leads, and practitioners with real-world Physical AI or robotics experience.
  • Status: Actively inviting industry speakers, reviewers, and mentors.
Join the project
Market research program

An industry-grounded research sprint focused on real Physical AI deployment constraints in Canada.

  • The Task: Conduct one in-depth industry interview focused on Physical AI deployment; extract and validate a single, high-impact pain point that blocks deployment, scaling, or ROI; synthesize findings into a standardized, one-page market-facing report.
  • Ideal for: Bachelor’s and graduate students in Mechanical Engineering, Computer Science, Robotics; early-career engineers interested in industry, startups, or applied AI deployment.
  • Status: Active research sprint with defined scope, interview criteria, templates, and publication pipeline.
Join the project
Metadata Forensics

DroneMeta is an open-source tool for extracting, normalizing, and validating metadata from consumer and industrial drone imagery—focused on reliability, provenance, and forensic integrity.

  • The Task: Parse metadata from DJI, Autel, Parrot, and open formats; normalize outputs into a deterministic JSON schema; detect inconsistencies, missing fields, and tampering indicators; build a public metadata “sample zoo” for testing.
  • Ideal for: Python developers, GIS / geospatial engineers, drone operators & forensic analysts.
  • Status: Definition complete. Development begins when the core contributor set is in place.
Join the project
Camera Diagnostics

DebugCam is a lightweight debugging tool for analyzing camera streams (USB, RTSP, ROS topics) and detecting dropped frames, timing jitter, and synchronization issues—without dashboards or GUIs.

  • The Task: Implement stream adapters (RTSP, USB, ROS image topics); measure latency, frame drops, and timestamp drift; output deterministic diagnostic reports; build reproducible tests for unstable streams.
  • Ideal for: Computer vision engineers, robotics developers, video systems engineers.
  • Status: Architecture and scope defined. Looking for maintainers to kick off v0.1.
Join the project
Failure Simulation

SensorReplay is an open-source toolkit for replaying and mutating sensor data to reproduce failure modes such as noise bursts, dropouts, drift, and desynchronization—entirely offline.

  • The Task: Design failure injection primitives (noise, dropout, delay); apply transformations to recorded sensor data; enable repeatable testing of perception and fusion pipelines; integrate with ROS 2 workflows.
  • Ideal for: Robotics testing engineers, simulation & validation engineers, safety-focused developers.
  • Status: Concept and use cases defined. Awaiting core contributors.

Do you have any other project in mind to collaborate? Let’s talk: contact@physicalai.ca

Why Contribute

What you will gain

Official Credit & Public Proof

Your name is listed on reports and assets distributed to government, labs, founders, and operators. It is permanent, searchable proof of contribution.

Verified Professional Endorsement

Upon completion, you receive a formal LinkedIn recommendation that captures your scope, execution quality, and results.

Titles That Signal Capability

You join as a Contributor, Associate, Research Fellow, or Sprint Lead—titles you can use immediately. No one is labeled a volunteer.

Network Access That Compounds

Work directly with founders, senior operators, and advisors shaping Physical AI. These are collaborators, not spectators.

Team-Building & Leadership Exposure

High performers are invited to form or lead sub-teams inside sprints, demonstrating ownership rather than participation.

Learning Through Execution

There are no lectures and no filler. You learn by shipping alongside peers operating at a high bar with real constraints.