Sr. AI Researcher, Agentic Systems
CA$181k - CA$230k • Canada • Full-time
Posted 2d ago
Job Location
Canada
Also open in U.S.
Tech Stack
Remote Work Policy
On-site
Employment Type
Full-time
Categories
AI Agent Engineer
About the job
OpenSpace is seeking a Sr. AI Research Engineer to build sophisticated agentic systems capable of reasoning across a vast array of project data, including 360° imagery, point clouds, BIM models, and technical documents. This role involves developing the layer above perception primitives, integrating frontier models to answer complex construction project questions and ensuring the answers are verifiable for practical application. The work is focused on applied research, tackling challenges in agentic reasoning, tool design, and ensuring high-confidence outputs for production deployment.
Responsibilities
- Design and build agent systems for multi-step task planning and execution across diverse data types.
- Develop and manage the context layer, including retrieval and grounding over indexed site records.
- Design the tool surface for agents, enabling them to locate components, compare captures, read drawings, and query BIM elements.
- Create and utilize evaluations for tasks with ambiguous ground truth to assess readiness for production.
- Select appropriate models (frontier APIs, open-weight, fine-tuned, classical CV) based on cost, latency, and accuracy.
- Ensure all outputs are traceable to their supporting evidence in the data.
- Prototype solutions rapidly, iterate on designs, and collaborate with platform engineers for production shipping.
Requirements
- M.S./Ph.D. in a related field or equivalent industry experience.
- 5+ years of experience building ML systems, with at least one LLM or VLM-based system shipped to production.
- Experience designing evaluations for data that is expensive, sparse, or contested.
- Strong multimodal understanding, including vision-language models, embeddings, retrieval, and context design.
- Proficiency in Python, async orchestration, tracing, and cost instrumentation.
- Comfort with spatial data concepts like coordinate frames, camera poses, and projections.
- Demonstrated judgment in identifying automation opportunities and assessing current limitations.