Subhajit Chatterjee

Software Engineer at Oracle NSGBU working on enterprise systems, production debugging, workflow automation, and previously on Generative AI systems. Interested in ML systems, distributed inference, GPU programming, and performance engineering.

About

I build software systems across enterprise products and AI infrastructure. My recent work involves production debugging, issue resolution, workflow automation, code/data fixes, and customer issue analysis for enterprise financial systems at Oracle NSGBU.

Previously, I worked on Generative AI systems, voice-based transaction workflows, zero-shot voice cloning, inference optimization using vLLM, and GPU-backed parallel processing.

Particularly interested in ML systems, distributed computing, GPU programming, and performance optimization.

Timeline

Mar 2026 - Present
Associate Software Engineer, Revenue Recognition Team (Rotation) - Oracle NSGBU
Rotated into Revenue Recognition module under Oracle's new grad rotation framework. Working on Escalation Resolution Communication (ERC): issue scrubbing, customer issue debugging, data fixes, code fixes, and product behavior investigations for Revenue Recognition customers. Worked on advanced intercompany journal QA tasks, ad-hoc migrations, and solved multiple U3 issues.
Sep 2025 - Feb 2026
Associate Software Engineer, Global NSNPA Team (Rotation) - Oracle NSGBU
Worked on Netsuite Global Payment Automation SuiteApp used to bunch sales orders, group requisitions/orders, and automate payments.

Contributed to API automation, UI automation testing, Jest code coverage, and TeamCity stabilization for flaky tests.

Built a workflow feature supporting CDL (Class, Department, Location) storage inside payment automation workflows. Stack: React, TypeScript, Jest, Jasmine.
Jun 2024 - Aug 2024
Project Intern (Generative AI) - Oracle
Developed a voice-based transaction system for purchase and sales orders, reducing inference latency by 7x and lowering costs through distributed processing with smaller transformer models, while improving parsing accuracy and integrating APIs for seamless transaction handling.
Sep 2024
DatSol Solutions
Developed a POC for zero-shot voice cloning, achieving 8x faster inference using vLLM optimization and A100-based parallel processing, with advanced acoustic and prosodic feature extraction for high-quality voice transformation.
Dec 2022 - Apr 2023
Informal Research
Under Dr. Puspanjali Mohapatra on swarm algorithms and neural networks. Implemented Extreme Learning Machine (ELM) with Salp Swarm Algorithm, achieving metrics similar to SGD-based models while reducing training time by 50%.
2021 - 2025
IIIT Bhubaneswar
B.Tech in Computer Science and Engineering, GPA: 9.30/10.0

Skills / Focus Areas

Backend Engineering • Production Debugging • React • TypeScript • Jest • Jasmine • TeamCity • Generative AI • LLMs • PyTorch • vLLM • GPU Computing • Distributed Systems • Workflow Automation • Performance Optimization

Projects

toy-bert 🔗

A rough reimplementation of BERT.

toy-gippity 🔗

A rough reimplementation of GPT (Causal LM) from scratch.

stone-grad 🔗

A tiny reverse-mode autodiff engine implementing backpropagation over a dynamic DAG.

SQLeen 🔗

An agentic framework that lets you query databases in natural language.

sweet-pea

A 1B LLM fine-tuned on simple Python programs with PEFT.

YeetCoin 🔗

A Bitcoin implementation built from scratch.

Publications

Imaging signatures for different mutation estimation for brain cancer 🔗