I design and productionize agentic AI frameworks, owning the full MLOps lifecycle from research to deployment.
Projects & Writing
Tunix: Reasoning Model Alignment
Distilled Chain-of-Thought reasoning into Gemma-3-1B-IT. Engineered a memory-efficient GRPO loop on a single TPU v5e using custom gradient accumulation strategies.
What's Good: AI Systems
Architected a distributed RAG pipeline on AWS, reducing p95 end-to-end latency from 17s to 6s through asynchronous parallelization. Implemented real-time vector interpolation for instant persona adaptation.
Curiocity: Multi-Agent AI
Engineered a 0-1 multi-agent conversational AI platform. Orchestrated agents via LangGraph and FastAPI with strict Pydantic schema validation. Fine-tuned local models with Unsloth/LoRA for specialized behavioral control.
About Me
I'm an AI Systems Engineer obsessed with the space where cutting-edge research meets production reality. My engineering DNA was forged in hackathons and leading technical communities, which taught me how to take abstract concepts—like reasoning model alignment or multi-agent orchestration—and forge them into robust, scalable architectures.
I don't just build wrappers. Whether I'm writing custom JAX training loops for TPUs, orchestrating stateful LangGraph deployments on AWS, or evaluating agent faithfulness with RAGAS, my goal is always end-to-end ownership. I believe the future belongs to engineers who can bridge the gap between algorithmic research and high-performance, fault-tolerant systems.
Experience
Associate Data Scientist (ML/AI Engineer)
Epsilon • Aug 2024 - Present
Architected and deployed production-grade agentic AI solutions, orchestrating multi-agent workflows from R&D to REST API model serving. Rigorously validated Text-to-SQL agents using RAGAS to ensure a Faithfulness score of 0.9. Built a multi-model propensity framework on Databricks driving a 15% conversion lift.
Researcher & Conference Speaker
12th ICBAI Conference • 2025
Authored and delivered an oral presentation on "Novel Hybrid GNN-LLM Recommender System via Behavioral Semantic Profiling." Delivered a 34.5% lift in NDCG@20 over LightGCN baselines in text-scarce environments using InfoNCE loss.
Data Analyst Intern
Epsilon • Feb 2024 - Aug 2024
Engineered a scaled implementation of the TIFUKNN research paper in PySpark, delivering a time-aware recommendation system that achieved a 7% lift in NDCG. Developed a projection modeling framework on a 10M+ user dataset.
B.Tech, Computer Science (Spec. in AI & Robotics)
VIT Chennai • 2020 - 2024
Graduated with a 9.09/10 CGPA. Served as President of the Zero Bugs Club, leading a 50+ member technical group and organizing hackathons for 500+ participants.
Core Competencies
Interested in building the future?
Get in Touch.
The best way to reach me is by email.