I build backends that hold under load. Distributed, event-driven systems in Python and Rust, at Hakimo. Off the clock, a passionate builder who keeps shipping products.
Based in Bengaluru. Open to opportunities and conversations.
I am a software engineer in Bengaluru. At Hakimo I build distributed, event-driven backend systems in Python and Rust.
Before that, data engineering at DataGrokr, and backend and frontend roles across India and Nepal.
Outside work, shipping is closer to a meal than an event. I get curious about a problem, build the smallest thing that helps, and put it in someone's hands. Nine products so far, two of them live.
Nine products I designed, built and shipped. Two are live and open to anyone. Each one lists what it does and the problem that was actually hard to solve.
Writing and voice notes become mood patterns, goals, reminders, and reflection worth reading back. Mobile app, responsive web, and the deployed API are all mine.
The hard part. Transcription and meaning extraction run on the user's own device, so a private journal never leaves it.
Friends answer from a single link with no accounts. Once enough answers land, a shareable AI verdict unlocks.
The hard part. Anonymity and one vote per person are opposite requirements. Private device identity, atomic duplicate protection, and guarded game states resolve it.
Prescriptions, reports, medicines, reminders, meals, family timelines, and lab trends live in one app. Every record pictured is fictional demo data.
The hard part. Health data cannot be shipped to a cloud model for parsing, so OCR, clinical entity extraction, and identity redaction all run on the phone.
Sathi builds the storefront, sells in Nepali conversation, recommends products, fills the basket, and takes cash-on-delivery orders.
The hard part. The shopkeeper never writes a product description or a reply. The model does the selling and the merchant only sees orders.
Sellers publish price lists. Retailers browse, build wholesale quantities, order, negotiate in chat, and get status updates.
The hard part. Retailers will not install an app to place one order, so the whole buying flow works from a link while the seller keeps a native app.
QR ordering, kitchen display, billing, staff, inventory, purchasing, recipes, shifts, and customer credit in one system.
The hard part. A bill must never be wrong, which is why this codebase carries 780 test cases across billing, inventory, and shift handover.
Leads, follow-ups, role-based access, and reporting for counselling businesses running several branches.
The hard part. Tenant isolation is enforced in the data layer rather than trusted to application code, so one institution can never read another's leads.
Any app, any screen, translated in place on the device for readers who are more comfortable in Hindi.
The hard part. Reads the live view hierarchy through Android accessibility nodes and renders translations as overlays, with no screenshots and no network round trip.
DNS, connect, TLS, and transfer time printed for interactive curl calls, with the response left untouched.
The hard part. Timing output has to reach the terminal without contaminating stdout, so piped and scripted curl calls behave exactly as before.
Jun 2025 — Present
Full-time · Bengaluru, India
Distributed, event-driven backend systems in Python and Rust.
Jan 2025 — Jun 2025
Internship · Bengaluru, India · Hybrid
Backend services, deployment pipelines and CI automation.
Aug 2024 — Dec 2024
Internship · Bengaluru, India · Hybrid
Data engineering on AWS: EC2, Batch, Lambda, PySpark jobs and Snowflake warehousing.
Mar 2024 — Aug 2024
Internship · Remote
Jul 2023 — Jun 2024
Community · Bengaluru, India · On-site
Taught Angular to student cohorts and built React projects with them.
Jul 2023 — Aug 2023
Internship · Remote
Node.js services.
Jan 2023 — Feb 2023
Part-time · Butwal, Nepal
Feb 2021 — Jul 2021
Part-time · Butwal, Nepal
What I reach for, grouped by what it does.
Jain (Deemed-to-be University)
CGPA 9.5
New Horizon College
Grade A+
Zero to one
A real product in users' hands in weeks, not a slide about one.
An AI feature that works
On device or hosted, wired into your product, evaluated, not demoed once.
A stalled build rescued
I read the codebase, find what is actually blocking, and ship the next release.
Bangalore, India