Building NepalIPMS at production scale
The Problem
IP firms in Nepal were running practice operations across spreadsheets, shared drives, personal calendars, and memory. Renewal windows closed without warning. Department of Industry status was discovered when a client called. The product had to become the desk — not another tab practitioners ignore.
Context
NepalIPMS needed to hold a national-scale trademark register alongside confidential client matters. Search had to work during live phone calls. Documents and bulletins needed audit trails. The team shipping and operating it was small — infrastructure complexity would kill the product as surely as a bad domain model.
Constraints
Confidential legal data. Downtime during filing season is unacceptable. Mixed Nepali/English mark text. Budget and ops capacity that rule out always-on server fleets. Practitioners will not reorganize their mental model around a generic CRM.
Architecture
Domain model: company → file → matter → asset. Next.js firm portal. Workers API with 300+ endpoints. D1 for relational state and FTS5 search across 130K+ records. R2 for filings and bulletins. KV for sessions and hot cache. OCR intake as a gated pipeline into the same matter system. See the architecture gallery and decision library.
Implementation
Shipped in slices aligned to desk workflows: matter records and deadlines first, register search next, then OCR bulletin intake with human review. 150+ database migrations tracked in version control. Every migration assumed production data firms depend on daily. Auth and role boundaries matched how associates and partners actually share files.
Result
Live at nepalipms.com. Used by IP firms including Global Law Associates and Janak Bhandari & Associates. Register search and deadline engines carry daily load. OCR reduces re-typing without bypassing expert confirmation.
What I Learned
Production scale in LegalTech is not vanity metrics — it is whether a paralegal trusts the screen while a client is on the phone. Domain shape beats horizontal abstractions. Edge infrastructure is a product decision when your team is small and latency is trust.