Build, ship, learn.
Notes from the team building Apragya AI. Product launches, engineering deep-dives, design decisions, and lessons from enterprise rollouts.
Editor's pick.
Why we built the agent runtime around LangGraph (and what we'd do differently)
12 months of running production agents on a graph-based runtime. The patterns that worked, the ones that didn't, and the architectural decisions we're betting on for the next 12.
Co-founder, CTO
Platform & AI
All posts.
Product
Industry pipelines: why we shipped Real Estate first
The reasoning behind choosing Real Estate, KYC, and Invoice Processing as our first three industry-tailored pipelines.
VP, Product
Jun 5Engineering
Building tenant-isolated multi-region from day one
How we architected for EU, US, and APAC region pinning without ever ripping out the data layer.
VP, Engineering
May 28Engineering
RBAC v3: rethinking enterprise permissions from scratch
Why we replaced our role/permission model after 14 months in production, and what the new system looks like.
Senior Engineer
May 19Design
Designing for AI-first interfaces (when AI is the OS)
Five UX patterns we've found that work when AI isn't a feature - it's the foundation users interact with constantly.
Design Lead
May 8AI Research
Picking the right model per task: a routing approach
How we route agent calls across Claude, GPT-4o, Gemini Pro, and self-hosted models based on task type and cost.
Applied Scientist
Apr 30Engineering
What "audit-first" actually means at the database layer
Every soft delete, every reviewer signature, every agent action - logged. The schema patterns we use to make it cheap.
VP, Engineering
Apr 15Customers
From 12 tools to one platform: a manufacturing tenant rollout
Detailed account of how a 400-person manufacturing org consolidated their AP, HR, and procurement on Apragya.
Product Lead
Apr 8Weekly notes from the team.
Get new posts in your inbox
One email a week, on Thursdays. Product updates, engineering posts, and customer stories.