Building to sell outcomes, not subscriptions.
Vihaya is an agentic decisioning platform for regulated enterprise workflows. The engagement model: embed with one customer for a 12-week pilot and ship a deployed AI service designed to take a meaningful share of their highest-volume decisions touchless, with every outcome audited, cited, and escalation-protected. The company is pre-revenue as of May 2026; the first design-partner pilots are the immediate roadmap.
What we are
An agentic platform plus a small team of engineers. We pair the two with one F500 buyer per engagement, ship into their environment, and leave the system with them. The platform is the substrate; the engagement is the product.
What we are not
- ✗ A SaaS subscription. You pay for a project; the system stays yours.
- ✗ A workflow builder. Citizen-developer drag-and-drop is a different category. Our buyer is the CDIO with a budget problem, not the ops lead composing flows.
- ✗ A horizontal “AI agent” product. We solve one vertical wedge at a time. Same engine, different domain expertise per engagement.
- ✗ AGI. The intelligence lives in the foundation models. Vihaya is the substrate that makes them safe to deploy.
How we work
Every engagement is a 12-week paid pilot focused on one specific workflow. Weeks 1–2 are discovery and access. Weeks 3–8 are ingest and integration. Weeks 9–11 run the system against shadow or parallel traffic. Week 12 is handoff. After the pilot, customers either renew into production-phase work or stop — there’s no lock-in.
Our principles
- Workload fit over fashion. We use Postgres + pgvector, not Neo4j. The decision is grounded in the actual query shape, not what looks good in a deck.
- Audit is a primitive. Every decision is one immutable row. Reconstructable from cold storage years later. SOC 2 evidence is a side effect, not a fire drill.
- Confidence floor is a primitive. Below threshold → human review, regardless of what the model said. There is no “100% automation” path; there is a meaningful-touchless-share path with a clean human-in-the-loop, tuned per pilot against real customer data.
- Evals block deploys. Every solution ships with a golden dataset and a CI-blocking eval gate. Drift below threshold → the deploy never reaches prod.
- Model-agnostic. OpenAI, Anthropic, Azure OpenAI, Google Vertex. Use whichever provider your security team has a BAA with.
Where we are today
Pre-revenue. Pre-SOC 2 Type II. Architecture is production-shaped; certification and HIPAA-BAA execution are on the near-term roadmap and we work transparently with security teams through pilots. We’re actively in pilot conversations with healthcare payors; the prior-authorization solution is the first vertical shipping.