Work

Things we've built that people actually use.

Four projects, three industries, one operating principle: ship the smallest thing that delivers real value, then earn the right to build more.

Durian Industries · In production

DurianOS — tender intelligence for a furniture giant.

Durian Industries is one of India's largest organised furniture brands — 1500+ employees, 65+ stores, six factories, and a big institutional B2B side selling to government, PSUs, hospitality and corporates.

That last bit is the catch. The Indian government posts thousands of furniture tenders a week across five different portals — GEM, CPPP, CWPD, IREPS, GePNIC — each with its own format, login, and quirks. Durian's sales team was drowning.

We built DurianOS. It scrapes all five sources across 28 states, normalises the data, runs a relevance engine to filter the noise, and routes the ~30 viable tenders per day to specific sales reps based on territory, product fit and past wins.

From there, the platform tracks the bid through the full lifecycle — Technical Evaluation, Financial Evaluation, Result — with email sync, document libraries, decision history audits, and a pipeline view.

13,158
Tenders ingested
5
Government portals
28
States covered
5
Sales reps using it daily
"It does in an hour what we used to do in a week." — Operating principle behind DurianOS
Miracle · Our product · Pre-launch

Miracle — invoice automation that lives on WhatsApp.

Every small Indian business spends real hours each week typing invoices from a paper or PDF into Tally. The vendor's "Cheeni" needs to become your "Sugar." The GST has to add up. The voucher has to land in the right ledger.

Miracle replaces all of that. Forward an invoice to the WhatsApp bot. Our AI extracts vendor, line items, amounts, GST, dates — then validates the math, catches the mistakes, syncs it into Tally as a real voucher.

Two non-obvious things that matter: it learns your item mappings (correct "Cheeni → Sugar" once and it's remembered forever), and it has a validation engine that catches the qty × rate mismatches and the GST errors before they hit your books.

Pre-launch now, with early access running. Read more at miracleai.in.

King Sarees · Shipped

ERPNext for Chandni Chowk.

King Sarees is a wholesale and retail saree house in Chandni Chowk — the kind of business that's been running on Tally and ledger books for two generations.

The brief: move the operation to ERPNext without scaring anyone. The team isn't anti-tech; they're anti-friction. If a new tool takes three more clicks than the old one, they go back to the old one.

So we redesigned the home dashboard from scratch. Collapsible sidebar sections — Purchase and Sales expanded because they're daily; Selling Setup and Masters collapsed because they're not. Three priority zones, biggest in the middle. Plain language, large tap targets, sections that get out of the way until you need them.

The Tally migration was its own project — an .intact backup masquerading as .mdb, six tables, full historical data preserved, mapped into ERPNext's schema.

In the lab

More to come.

A few projects we're not ready to talk about yet — including the next thing we're building with Durian, the v1 launch of Miracle, and at least one workstream we've quietly started for a new client.

If you're a builder who wants to own one of these, our careers page is where to look.