The company story
Athyna Insight (formerly Inplico) set out to make Canadian building code compliance faster without pretending regulation is a toy. As CTO and co-founder, we took the product from concept to production for Edmonton’s construction industry — with licensed experts validating the work.
Not hours of code lookup
Seconds
Validation
Expert-in-the-loop
Lookup time
Hours → seconds
Role
CTO / co-founder
Athyna Insight — impact snapshot
The pressure
What was on the line
Building code interpretation has to be accurate, reviewable, and scalable across municipalities. Practitioners were spending 4–6 hours a week on lookups across large multi-document regulatory corpora. The product had to earn trust from Safety Codes Officers and architectural technologists — not just impress a demo audience.
What changed
The arc of the work
- 01
Design for expert validation
Human-in-the-loop validation with Safety Codes Officers and architectural technologists was part of the product, not an afterthought. Confidence scoring and review paths kept professionals in control.
- 02
Make the retrieval stack production-grade
Hybrid vector-metadata search and document processing pipelines brought common address-to-guidance lookups from long manual research toward seconds — grounded in the actual regulatory corpus.
- 03
Ship for a real construction market
Concept to production deployment serving Edmonton construction, with partnerships adjacent to National Research Council task group members shaping 2028 building code policy. Technical leads recruited including a Kaggle top-851/240K profile.
Product in the wild
AthynaInsight.ai — real output
Design prototype — expert feedback with pinned code references. Greyscale to match this site’s palette.
Before and after
What people could feel in the week
4–6 hours a week on building code lookups
→Answers in seconds, grounded in the regulatory docs
Accuracy hard to verify across large code corpora
→Expert-validated, citation-linked answers for review
Where they are now
The outcome that stuck
A compliance product that keeps expert judgment in the loop — address-to-guidance responses in seconds for common lookups, with citation-linked answers designed for professional review.
For technical readers
How it was built
Hybrid vector-metadata search, LangChain + Gemini document processing, Google Cloud (Vertex AI, Document AI, Cloud Run), BigQuery vector search, expert validation, dynamic zoning intelligence, Next.js, Terraform.
GCP Vertex AI · Document AI · Hybrid RAG · Gemini · Next.js · Terraform
More company stories
- CopperTeamsCopperTeams’ 20% team-performance guarantee — extended with AI coaching between human sessions.
- A $26Bn Hedge FundAgentic research validation in a $26Bn portfolio context — days of gather compressed toward hours of focus.
- A Canadian Government AgencySiloed housing archives became searchable — with high-confidence document classification in a two-phase POC.