Open to Opportunities

Koustubh Kulkarni

Software Engineer

I engineer reliable, scalable systems - from backend APIs to applied AI.

About

Computer Science undergraduate at KLE Technological University (CGPA 8.7/10), graduating 2026. I've interned at NVIDIA and KPMG, building everything from virtual camera systems and HDR pipelines to Agentic AI systems and LLM-powered products. I thrive at the intersection of backend engineering and applied AI - designing systems that are fast, reliable, and actually useful. Outside work: ACM ICPC Regionalist, CodeChef 3-star, and a competitive problem solver with 400+ DSA problems solved. I care about clean architecture, clear thinking, and shipping things that matter.

Applied AI & LLMsBackend SystemsDistributed ArchitecturesComputer Vision

Experience

Digital Lighthouse Intern

KPMG - Bangalore, India

Jan 2026 - Present

-Engineered Agentic AI systems using LangChain & RAG integrated with Snowflake for enterprise data retrieval.
-Built a modular LLM Presentation Generator with LangGraph and pptxGenJs - workflow-orchestrated, cost-optimized inference.
-Trained in enterprise-scale Java, Spring Boot, and cloud-native application design.

Project Intern

NVIDIA - Hubli / Remote

Aug 2024 - Jun 2025

-Built virtual camera systems using VirtIO for real-time streaming between Linux and Android VMs, reducing streaming latency by 30% via optimized QEMU and FFmpeg pipelines.
-Developed an HDR reconstruction pipeline using a dual-branch U-Net for multi-exposure image fusion, trained on 60K+ SDR and IR images.

Projects

Featured Builds

Hack Karnataka 2nd Place

CreekPay

Blockchain BNPL Platform

AI-powered P2P Buy Now Pay Later system with AI-driven e-KYC, ML-based credit scoring, and real-time credit evaluation. Won 2nd place at Hack Karnataka among 100+ teams.

FastAPIPostgreSQLOCRMachine LearningBlockchain

Role: Led AI/ML development and backend architecture in a 3-member team.

GitHub2025

FinMitraAI

AI Financial Advisor

LLM-powered financial advisor delivering personalised investment insights with 85% accuracy. Processes 1,000+ financial data points in real-time via market analytics pipelines.

PythonFastAPILangChainGroq APIs

Role: Full-stack solo build - LLM integration, backend, and pipeline design.

GitHub2025

Synk-App

Real-Time Chat Application

Secure WebSocket-based messaging platform with JWT authentication, persistent MongoDB storage, concurrent session handling, and scalable backend event management.

Node.jsSocket.IOMongoDBJWT

Role: Solo build - architecture, backend, and auth system.

GitHub2025

GitHub Repositories (koustubh-k)