39 / Legal
On-Premise Legal Intelligence
We built a fully local legal search index: over a million pieces of Indian legislation plus the client's own case files, running on our own GPUs, so privileged documents never touch the cloud.
- Client
- Real Estate Developer's Legal Team, Plus 360 Labs' Own Indian Law Index
- Stack & approach
- Legal AILocal InferenceBGE-M3 EmbeddingsIndic OCROn-Prem DGX
The problem
Legal teams need fast search across Indian law and their own case files without sending privileged documents to the cloud. The client's existing setup took 15 to 20 minutes per query.
Solution & architecture
A fully local legal search index: BGE-M3 embeddings combined with keyword search over Indian legislation and judgments, running on our own GPU hardware. An OCR extraction pipeline handles scanned case files including Devanagari documents, and a lease-deed extraction model reads 19 fields and 8 red flags, all on-premise, with vLLM serving for faster embedding throughput.
Why it is built this way
Privileged files never leave
Local inference on our own GPUs
No per-query cloud cost
Embeddings served locally with vLLM
Indic-capable
OCR for Devanagari and mixed-script
