Freelance product developer.
Hyderabad, IST.
Replies within a working day.
I build web products end to end.
Full-stack SaaS, AI features and frontend-heavy apps, for founders, small teams and agencies. The three products below are live, open source and measured.
Work
Docwise
AI extraction of supplier invoices, with every value audited. Claude reads each invoice into a strict schema. Rules, the model’s own doubt and the document’s text decide what is approved and what waits for a person.
On 20 held-out invoices, scored once: 20 fully correct, 0 wrong values marked high confidence, and a median model cost of $0.0115 per invoice.5
Slotwise and Formwise connect: Formwise forms are the intake questions in Slotwise, and both apps check the answers with the same code.
How I work
Small scope, working software every week.
Scope call
We talk for 30 minutes about what you need. Within 2 working days you get a written scope and a fixed price.
Weekly deploys
Every week you get a link to the working app, not a status report. You click through what’s done and tell me what to change.
Handover
You get the code, tests that run on every push, and a README on how to run, deploy and change it. The repos above show what that looks like.
Price
Three ways to work together.
Feature sprint
$1,800
per week
One feature or fix, scoped up front, built and deployed in a week. A good first project.
Start here if we haven’t worked together.
MVP
from $6,000
fixed price
A first version of your product, end to end, usually 3 to 4 weeks.
Retainer
$3,500
per month
About 20 hours a week of ongoing development on your product.
Not sure what you need? The scope call is free.
Contact
Have something to build?
Email me what it is and when you need it. Replies within a working day.
krishna@bhupathi.devSources
- Slotwise booking page, Lighthouse mobile: performance 98, accessibility 100, LCP 2.3 s. Median of 3 runs, Lighthouse 12.8.2, measured 14 Oct 2026 from India against production. Source: slotwise-booking/docs/measurements.
- Slotwise tests as of 15 Oct 2026: 156 Vitest tests (unit, and database tests on PGlite) and 18 Playwright runs (9 tests on desktop Chrome and a Pixel 7), green in CI. Source: slotwise-booking README.
- “Describe your form” on Claude Opus 5.5: 8 of 8 descriptions returned a valid form. End to end from India to the US region, median 4.45 s, range 2.91–7.67 s. Measured 15 Oct 2026. Source: formwise-builder/docs/measurements.
- Formwise builder, Lighthouse mobile: performance 98, accessibility 100, LCP 1.8 s. Median of 3 runs, Lighthouse 12.8.2, measured 14 Oct 2026 from India against production. Source: formwise-builder/docs/measurements.
- Docwise held-out split: 20 generated invoices, scored once after tuning ended, with prompt v1 on Claude Sonnet 5.5. Header fields 100% correct, line-item F1 100%, 20 of 20 fully correct, review rate 5% (1 of 20), 0 escaped errors. Model cost per invoice at list prices: median $0.0115, p95 $0.0170. Time per invoice: median 3.1 s, p95 5.6 s. Source: docwise-extract/docs/measurements.


