Computer Science @ PES University · Bengaluru
I build systems at the intersection of AI, systems, and infrastructure.
Currently exploring:
- 🧠 LLM inference, post-training, and model efficiency
- ⚡ KV-cache compression and efficient inference
- 🤖 Agentic AI and multi-agent systems
- 🗄️ Database internals and distributed systems
I like working close to the underlying systems, where performance, correctness, and reliability actually matter.
LAKV · Adaptive KV-Cache Compression
A training-free KV-cache compression framework for Qwen2.5-7B-Instruct.
→ 2.81× KV-cache compression
→ Layer-aware importance scoring
→ INT8 / INT4 / token dropping
→ Investigating compression vs. reasoning accuracy
| Project | What I built |
|---|---|
| LAKV | Adaptive KV-cache compression for multi-agent LLM systems |
| Logickos | Deterministic reasoning verifier using graph-based analysis |
| AlpacaDB | Relational database built from scratch with B-Trees and query optimization |
| SENTINEL | Multi-agent GenAI system built during my internship at Visa |
Visa · Software Development Intern
Jun 2026 – Aug 2026
Built SENTINEL, a 6-agent GenAI system that converts Jira stories into production-ready PySpark validation scripts.
- Schema-grounded routing across 7 enterprise data domains
- Deterministic business-rule extraction and code generation
- Static analysis + AST-based code repair
- 18 modular validation patterns
- React + FastAPI review interface
Mstack Chemicals · Backend Intern
May 2025 – Aug 2025
Designed and shipped a FastAPI microservice with 156 REST endpoints across 25+ PostgreSQL models.
Languages
Python C Java JavaScript SQL Dart
AI / ML
PyTorch HuggingFace Transformers LangGraph LLM APIs OpenCV
Systems
PySpark Kafka Trino Docker Kubernetes
Backend
FastAPI Flask PostgreSQL MySQL MongoDB
Building things, breaking things, figuring out why.