Brief description of the organization
EcoWeaver is an organization developing decision support tools for ecological restoration and conservation practitioners. We have created an open knowledge framework linking causal ecological evidence to restoration practice, built on the Causal Mosaic Schema (CAMO) — a schema co-developed with philosophers of causality for documenting not just that X causes Y, but how, and how confidently. It's developed out of a postdoctoral translational-ecology research program spanning the Bennett Lab (Carleton University) and DRAGEN Lab (University of Waterloo).
Problem area
Restoration practitioners need trustworthy answers to practical questions — "does removing buckthorn help oak regeneration?" — grounded in the actual evidence base. General-purpose LLMs, and even standard RAG systems, are prone to filling gaps with confident-sounding fabrication, and rarely know how to say "we don't know" gracefully or trace an answer back to its source.
Main objectives
- Build an agent or RAG system that answers practitioner questions using only a CAMO-structured knowledge graph (provided) as its evidence source — never inventing claims the graph doesn't contain.
- Ensure every answer is traceable to specific causal claims and source papers, with calibrated confidence.
- Evaluate the system against realistic practitioner questions for usefulness, traceability, and — just as important — how gracefully it handles questions the graph can't actually answer.
Scope of work
- Review of the CAMO graph schema and knowledge graph
- Design of a retrieval/reasoning architecture (graph-grounded RAG, agentic tool use, or a hybrid)
- Engineering for calibrated uncertainty and source provenance in every answer
- User testing with practitioner-style questions
- Documentation
Deliverables
- Report
- Website
- Resources
- New protocols/processes
- Presentation
- A delegate from the University of Waterloo will have the opportunity to present at an international conference in Germany in March 2026. (potential for travel subsidy; or virtual)
Team meeting frequency
Weekly
Skills and training required
- Python
- LLM/agent frameworks
- Some experience querying knowledge graphs (SPARQL, Cypher, or graph libraries)
- Ecological literacy is helpful but trainable
- A grounding in epistemology or philosophy of science is an asset for this project specifically, given how central "knowing what you don't know" is to the brief.
Resources required
- Access to Nibi/GPU compute (some provided)
- The pre-built CAMO graph (provided); LLM API access.
NDA or a commercialization agreement for this project?
Yes