AI/ML Product Engineer · Researcher · Builder

I build AI capabilities for real products — from classical machine learning and anomaly detection to retrieval systems, LLM applications, agents, evaluation, and inference-time reasoning.

I'm currently an AI/ML engineer at Oracle Financial Services Software, working on bringing machine learning into enterprise products. My strongest professional experience is in ML product engineering: model/backend implementation, SQL and application integration, reusable frameworks, validation, performance investigation, deployment support, release readiness, and technical architecture.

Outside work, I research modern AI systems, build open-source projects, write about ML, and help run the Gradient Ascent Syndicate.

• ML product engineering — anomaly detection, predictive modeling, explainability, evaluation, integration, performance work, and release-oriented engineering
• LLM systems — RAG, document intelligence, code generation, prompt engineering, and inference-time improvement
• Agents & evaluation — agentic workflows, tool use, reliability, and evaluation methodology
• Applied AI research — especially reasoning, code generation, knowledge graphs, and efficient inference
• Open-source education — building practical, implementation-first ways to learn AI/ML

A multi-source biomedical reasoning system that connects drug mentions to adverse events using CADEC, RxNorm, OAE, SapBERT, FAISS, and knowledge-graph path tracing.

A modular, provider-agnostic RAG system supporting local and cloud LLMs, FAISS retrieval, document indexing, and a LangGraph-based agentic workflow.

HumanEval experimentation across self-consistency, structured reasoning, self-planning, self-refinement, recursive criticism, progressive hints, and test-based enhancement.

An open initiative to build a comprehensive, implementation-first AI/ML curriculum, from mathematical foundations and classical ML through deep learning, modern LLM systems, and agents.

→ View the curated project portfolio

My research interests sit around one central question:

How do we make AI systems more capable, reliable, and useful without simply throwing a larger model at the problem?

That has taken me through:

• prompt engineering for code generation
• inference-time amplification and iterative refinement
• knowledge-graph-assisted biomedical reasoning
• LLM evaluation and code-generation benchmarks
• agent architectures and agent evaluation

I also have work published at IEEE ICDH 2026 on drug–adverse event reasoning using knowledge graphs.

Agents · Agent evaluation · Inference-time compute · Reasoning systems · Reliable enterprise GenAI · AI product architecture

I'm particularly interested in the engineering layer between "the model works in a notebook" and "the capability is integrated well enough to become a real product feature."

I also write, draw, make things, and occasionally get on a stage with a microphone. I like the idea that technical depth and creative expression do not have to belong to different people.

_{Hyderabad, India · Building at the intersection of AI research, engineering, and products.}