My name is Soumyajit.


I’m a software engineer with a strong foundation in data science, blockchain, and AI-driven systems. I’ve worked on healthcare tech at Carelon Global Solutions, where I built LLM-powered pipelines for processing clinical documents and deployed explainable AI tools in production. As a freelance engineer at a stealth startup, I’ve implemented smart contract logic and full-stack DApp workflows on Ethereum-compatible chains. I’m interested in building intelligent systems, shipping resilient infrastructure, and working at the intersection of emerging technology and real-world impact.

I like to think deeply about the design of software products. Like haute couture clothing, software should be elegant, functional, and effortlessly designed. I believe great software becomes a seamless part of daily life — a habit, and eventually, an extension of the mind.

The ultimate purpose of any software system is to make the user’s life easier. It shouldn’t demand mental effort, but rather ease cognitive load and quietly support the user’s intent.

Outside of work, you’ll probably find me running or lifting — both keep me sane and are as essential to me as food or sleep. I love traveling, diving into rabbit holes about science, history, and international politics, and geeking out over great books. Oh, and coffee? I’m obsessed — always experimenting with new ways to brew the perfect cup.

I was born in Guwahati, a small city tucked away in the northeast of India. I stayed close to home for college, graduating in 2022 from IIIT Guwahati with a B.Tech in Computer Science and Engineering.

The Road So Far

2024 – Present
Software Engineer (Freelance)
Stealth Startup
  • Developed AMM and flash loan logic in Solidity for a DEX platform.
  • Achieved 98%+ test coverage for financial transactions.
  • Built RESTful/Web3 APIs, reducing client-side failures by 25%.
  • Integrated Web3.js workflows in a React-based DApp.
2022 – 2024
Associate Software Engineer
Carelon Global Solutions
  • Built LLM-based RAG pipeline for EMRs; boosted decision accuracy by 85%.
  • Created tokenizer-backed document chunking for large-scale retrieval.
  • Designed prompts to extract structured data; improved F1 score by 22%.
  • Implemented prompt chaining with fallback for LLM robustness.
  • Aligned deployments with HIPAA via compliance collaboration.
2022
Data Science Intern
Carelon Global Solutions
  • Researched SHAP, LIME, and Captum for XAI in clinical ML workflows.
  • Deployed model explainability as FastAPI microservice.
  • Reduced override rate by 24% via integrated model rationale.
  • Enhanced error analysis with token-level attribution.

Get in touch

I’m most active on Twitter and GitHub. The best way to contact me is either through Twitter, or my email at debsoumyajit@gmail.com