About Dvar
An evaluation-led approach to enterprise AI
01 · Approach
Quality is measured first; the lowest-cost model that meets the bar is then deployed.
- 01 Measure5 minutes to connectEvaluate every response
Quality checks are defined with the client team, so quality is quantified.
- 02 CompareUp to 50× lower costSelect the right model
Open and frontier models are compared on the same checks, for quality and cost.
- 03 RunHosted in IndiaServe in production
A single API for every workflow, hosted in India.
- 04 TrainUp to 40× lower cost than open modelsCustom training
Where a quality gap persists, a small open model is trained and owned by the client.
02 · Team and clients
Leads
- Aditya BansalDa Vinci. IIT Delhi.
- Jayesh JawandhiaBain Capital, BCG. IIT Delhi.
- Chirag BhattMicrosoft CoreAI. IIT Delhi.
- Rehan BhatiaMcKinsey. IIT Delhi.
- Aman SepatGoogle, Glean.
Broader team from
- Hosted in India
- DPDP Act 2023India
- SOC 2 Type IIAudit under way
- ISO 27001Audit under way
- VAPTSecurity tested
- GDPREU
Clients
- LendingSecond-largest gold loan company
- InsuranceLeading insurance marketplace
- LendingProperty-backed lender
- LendingConsumer lending company
- BrokingLarge stockbroker
- ConsumerMicrodrama platform


