A staged view of the build — where we are today, what we are hardening, and the road to launch.
Iterative testing of the core engine across civil, criminal, and custody filings to make sure the platform handles real-world document variations smoothly.
Running ground-truth validation against real Supreme Court filings, working toward the launch accuracy target.
A working beta placed in the hands of design-partner Advocates-on-Record, running real, consented filings pre-submission.
Public availability for Advocates-on-Record — the expected launch milestone.
* Tough, but not impossible.Built on a proprietary dataset of real Supreme Court filings.
Our rule engine is built on a proprietary dataset of over 2,000* SLPs filed in the last 8 years, capturing more than 10,000* unique defect instances. The rule knowledge base is now locked, and the detection engine is validating against real Supreme Court filings — both digital and scanned — across civil, criminal, and custody matters.
Special Leave Petitions in the proprietary dataset.
Span of filings the dataset draws from.
Unique defect instances captured across the dataset.
The hard, honest work between here and launch.
Estimated accuracy target at launch, measured against ground-truth Supreme Court filings.
The detection engine now runs against real Supreme Court filings — both digital and scanned — while we harden error handling across every detection surface and run ground-truth validation, working toward our 78–82%* accuracy target at launch.
A subscription model with two tiers for Advocates-on-Record (AoRs) and regular lawyers, with plans to expand horizontally across major High Courts.
Expanding to a high-volume platform serving 4,00,000+ undertrial prisoners and their families, providing free document understanding and clear case summaries.
Connecting consumers to verified counsel based on objective performance data like bail success rates.
Progress state updated as of 30 June 2026; may not reflect live progress. For real-time updates, reach out to vaibhav@nyave.in.
* Based on our own dataset; actual figures may differ.
