The environmental health intelligence company. AI-first. Evidence-grounded. Built in the open.
vasus.ai exists because the science of how environment affects health is clear, the data to act on it is now available, and the AI to connect them intelligently finally works. We built the platform to prove all three.
The gap was obvious. The platform to close it wasn’t.
70–80% of chronic disease burden is attributable to environmental exposures. That figure appears in the Global Burden of Disease study, in the exposomics literature, and in every major review of non-communicable disease. It is not a fringe claim. It is established science.
And yet no tool existed that took a person with migraines, asthma, or a cardiovascular condition and told them — with evidence — what their environment was doing to them right now. Weather apps gave numbers. Air quality apps gave indices. None connected those numbers to specific conditions, specific populations, or specific peer-reviewed evidence.
vasus.ai was built to close that gap. The exposomics science existed. Google’s Environmental APIs made real-time data available at scale. And the combination of semantic search over a large scientific corpus with grounded LLM synthesis finally made it possible to turn environmental data into something clinically meaningful and evidence-traceable. The timing was right. We built it.
“To make the relationship between environment and human health legible — for every person, in every place, grounded in evidence.”
Not AI-enhanced. AI-first.
vasus.ai was conceived, built, and is operated as an AI-first platform. This is not a traditional rule-based system with an AI wrapper. AI is the architecture — from development to data pipelines to the intelligence layer itself.
The platform was built using Claude as the primary development tool.
From architecture to implementation — the data pipeline design, EHSPI scoring, RAG retrieval logic, API layer, and frontend — Claude was the co-developer throughout. This is not an AI feature. It is how the entire platform was built.
Every structural decision and architectural trade-off was reasoned through in collaboration with Claude. The codebase is coherent, documented, and testable because it was built the way good software should be built.
Data pipelines managed by agentic AI with deterministic guardrails.
The full pipeline — tile refresh, feature computation, EHSPI scoring, corpus ingestion, alert monitoring, and the daily system digest — runs as a set of coordinated agentic Cloud Run jobs. Each operates autonomously within strictly defined deterministic guardrails.
Guardrails include: schema-validated outputs, idempotent write operations, structured observability logging with X-Request-ID tracing, and hard failure modes that alert rather than silently degrade. Autonomy with accountability.
LLM-driven recommendations grounded in peer-reviewed evidence.
The synthesis layer uses OpenAI gpt-4.1-mini to generate structured intelligence outputs. Critically, it cannot hallucinate: every recommendation is grounded in papers retrieved from 350,000+ peer-reviewed articles by the RAG pipeline before the synthesis prompt is even constructed.
The LLM’s role is synthesis and communication — not knowledge. It has no access to training data on this topic; it only sees what the retrieval pipeline provides. Grounding is structural, not instructional.
Imiya Abhaya Iriyagolle
BSc (Psychology & Neuroscience), MA (Strategic Marketing)
Imiya is a digital transformation leader and solution architect with a scientific background and 15 years of experience building and deploying data-driven technology platforms across global organisations including NEOM, Sainsbury’s, Vitality Insurance, Stanley Black & Decker, GSK and Reckitt Benckiser. His career spans CRM, AI, data platforms, and enterprise architecture across health, insurance, retail, finance, and luxury real estate — with a consistent track record of delivering measurable real-world outcomes in complex, high-stakes environments.
His academic background in psychology and neuroscience, combined with deep technical expertise in AI, data systems, systems thinking and API architecture — alongside childhood experience with asthma, allergies, and migraines — positioned him to identify a gap that sits at the intersection of environmental science, biomedical evidence, and personalised digital health. That gap — the absence of any platform connecting real-time environmental conditions to individual chronic condition management with scientific rigour — became vasus.ai.
vasus.ai is built on the emerging science of exposomics — the study of environmental exposures and their cumulative health effects — operationalised as a real-time intelligence platform. Imiya leads product, architecture, and commercial strategy, bringing together his experience in enterprise-scale data systems, agentic AI, and health technology to build what he believes is a fundamentally new category: precision environmental health intelligence.
Live. Real data. Real infrastructure.
Advisors & collaborators
We are in active dialogue with leading environmental health institutions and digital health operators. Named advisors and partnerships will be listed here when formally confirmed.
We’re building something real. Come be part of it.
Five founding partner slots. Direct product input. 50% of commercial pricing for the first year.
5 slots. Full API access. Direct input into product direction. 50% pricing for year one. For platforms serving users with chronic conditions.
Register interest →See what we built. No account, no install — try a real query and see exactly what the API returns.
Open the app →All weight vectors published. Full EHSPI methodology. 20-city coverage. Live scores. Nothing hidden.
View research →