$ cat projects/hims-agroha.mdx
HIMS Agroha — Hospital Information Management System
Hospital management system used daily across a 500-bed hospital. Patient-journey modules, an automated payroll engine that replaced 40+ admin-hours/month, and 20% faster APIs.
React · JavaScript · Redux · Node.js · MongoDB · MUI · EJS · 2022 · live
Daffodil Software · May 2022 – Feb 2023
HIMS Agroha is the hospital management system for a 500-bed hospital — the kind of software that doctors, nurses, pharmacists, and admin staff actually open in the morning to do their jobs. Stakes are real. Bugs get noticed.
what I built
- Hospital management system — OPD, IPD, Billing, Inventory, Pharmacy. The full patient journey, used daily by medical staff across a 500-bed hospital.
- Spearheaded Inventory Management, Billing, and IPD as the lead developer for those modules — owned the data model decisions and the integration points with the rest of HIMS.
- Automated payroll engine for 300+ hospital staff. Designed and owned the rule engine that encodes Haryana government pay-matrix rules — grade pay, allowances, deductions, the works. Replaced a manual Excel-based process that consumed 40+ hours of admin time each month.
- API performance — identified slow diagnostic workflows as a bottleneck for medical staff. Redesigned Node.js aggregation queries, added compound indexes, introduced query caching → reduced average API response time by 20%.
- Mentored 2 junior developers — assigned tasks, reviewed code, guided their growth. Both moved to independent projects within months based on their progress.
- Documentation — wrote end-to-end platform flows and onboarding guides, reducing ramp-up time for new team members.
what stuck with me
- Healthcare software is a humbling teacher. You can't handwave error states when a pharmacist is staring at the screen waiting to dispense.
- "Improving an existing system by 20%" beats "ten percent of a greenfield" most of the time, if the existing system is doing real work.
- A pay matrix encoded as code, with tests, is a genuinely beautiful artefact. The Excel sheet it replaced was the opposite of that.
what I'd do differently
- I'd write integration tests for the payroll calculations from day one. They were correct, but proving they stayed correct as the pay matrix changed needed more than manual spot-checks.
- I'd advocate harder for proper observability on the API tuning work. We measured improvement in response time, but a slow-query dashboard would have caught the next regression before a doctor did.