Lead Engineer at Chainrisk Labs. I lead a team of five across a risk engine, a simulation pipeline, and the cloud platform underneath them.
Most of my work is deciding what a system actually has to guarantee, then removing everything that isn't that.
Proof of Reserves over ~$1.5B in assets. Attestations are computed inside AWS Nitro Enclaves, so execution is hardware-isolated and independently verifiable, then proven with zero-knowledge circuits written in Circom — reserves can be verified without exposing position-level data. Since extended to real-world assets (RWA). → Write-up
A risk engine that costs 60× less to run. Rewrote Monte Carlo simulation in Rust with CUDA-accelerated kernels, replacing interpreted per-scenario execution with batched GPU compute: 60× lower cloud spend and 40× faster end to end. Parallelising Solidity test-case execution across cores took runs a further 16× faster.
6M+ large-scale simulations a day. Designed Chainrisk's cloud architecture from the ground up and owned it end to end — Docker, Kubernetes, Terraform and Ansible on AWS, every environment defined as code.
A stack that costs 30× less to operate. At DryFi, brought the software in-house on GCP and Firebase, taking monthly running cost from ₹30,000 to under ₹1,000.
| Project | |
|---|---|
| flexauth | Flexible, blazingly fast, secure in-house auth you deploy on your own servers. Rust Axum Tokio MongoDB |
| freeflow | A CreativeOps tool for the design industry — manages the creative process from brief to delivery. Next.js TypeScript GCP Firebase |
| mamavault-backend-go | Fully serverless backend for Prega, a document manager for expectant mothers. Go Firebase Cloud Functions |
| portfolio | My site — and a template anyone can fork. Next.js TypeScript Notion as CMS |
I write about the parts of infrastructure that don't make good conference talks — at rajdeepsengupta.com.
- Proof of Reserves for $1.5B: attestation inside Nitro Enclaves, verified with Circom
- Terraform Remote State Management with S3 and DynamoDB
- Building a Production-Ready Retrieval-Augmented Generation (RAG) Application
- Envelope Encryption: Securing User Data with Layered Protection
- Design the failure paths first — the happy path takes care of itself
- Cloud cost is a design constraint, not a monthly surprise
- Optimise CPU utilisation, not instance count
- Prefer simple systems that scale over clever ones that don't
Rust Go TypeScript · AWS GCP · Kubernetes Docker Terraform Ansible Linux · CUDA Circom Nitro Enclaves · PostgreSQL ClickHouse
Portfolio · LinkedIn · X · rajdipgupta019@gmail.com
Based in Kolkata, India (IST · UTC+5:30). Open to conversations about senior and lead infrastructure roles — remote, flexible across time zones.