The systems I integrate say 200 OK when they fail. I build the parts that notice.
I lead backend at Mulltiply.AI, a B2B commerce platform. Orders, catalogue, payments and every integration around them, on Node.js and Postgres.
A way into the PC in the back of the shop
A Tauri desktop agent that dials out through a tunnel, so the cloud can reach Tally and BUSY without the shop opening a single port.
Two ERPs that answer 200 OK when they fail
Two-way sync through a 6,000+ line TDL add-on and a reverse-engineered BUSY client, and now purchases and sales on a ledger shaped like Tally's.
POST localhost:9000HTTP/1.1 200 OK <RESPONSE> <CREATED>0</CREATED> <ERRORS>1</ERRORS> <LINEERROR>...</LINEERROR> </RESPONSE>
When Postgres quietly picks the slow plan
281 ms to 29 ms, 1.8 s to 2 ms, and a store delete from a day to 11 minutes. Each one proven before it shipped.
before · 1,802 msNested Loop Hash Join rows=6.8M Seq Scan whole catalogue Index Scan items
after · 2 msNested Loop Index Scan sku_id = ANY($1) rows=20 Index Scan items
Many stores behind one WhatsApp number
A visibility and registration matrix, illegal pairs refused server-side, and zero change for every single-store customer.
inbound messagenumber -> subscription config cached 5 min config -> visibility × registration illegal pairs refused retailer -> their store(s) cached 30 min any failure -> single-store behaviour the old path
Writing
Every post is about something that broke in production, and what it took to find out why.
-
Too slow is too big
My fallback for large Tally exports waited for a "too big" error. The biggest company never sent one. It timed out at Cloudflare's 120 seconds instead, and the fallback never ran.
-
The floor was the call
An LLM classifier took three seconds to answer even "confirm". Shrinking the output barely helped, because the model's fixed per-call cost was the floor. A faster model was right there, and I still didn't take it.
-
Five agents, three human approval gates
A multi-agent LLM system that spends real money on ads: every ad created paused, three human approval gates, and every model call logged.
Experience
The short version. The resume has the rest.
-
Engineering LeadMulltiply Tech India
Lead a team of 3 engineers. Primary maintainer of the core API and the order engine behind it, across 8+ Node.js services on Kubernetes on GCP. Built the Tally and BUSY integrations, the desktop connector, and our ONDC buyer app as primary author. Largest single contributor to a 16,500-commit core API.
Feb 2026 to now -
Full Stack EngineerMulltiply Tech India
Built the WhatsApp commerce flows on the Meta Cloud API (onboarding, ordering, catalogue, payments) and n8n automation for 10+ operational processes.
Jan 2025 to Feb 2026 -
FounderSoloNodes.net
Ran a profitable game-server hosting business while still in school: billing, automated provisioning, a 10,000-member community. Sold it in 2022.
2020 to 2022 -
BCA (Hons.), Data SecuritySICSR, Symbiosis, Pune
CGPA 8.61. Won the institute's internal hackathon in December 2024.
2022 to 2025
backend
Node.js, TypeScript, Python, REST, GraphQL, webhooks, Tauri desktop apps
data
PostgreSQL, Redis, DragonflyDB, MongoDB, Meilisearch, Elasticsearch
infra
Kubernetes, ArgoCD, BullMQ, Docker, Cloudflare Tunnel and R2, Datadog
integrations
Tally (TDL), BUSY, Meta WhatsApp and Instagram APIs, ONDC, OpenAI
Hiring for backend or platform work?
Looking for a senior backend, platform or engineering lead role. Bring me the integration nobody wants to touch. I reply within a day.
5,573 commits in 2026, incl. private 9,938 contributions since 2020 3,835 commits to the core API, the most of anyone