Current Focus

• Designing production-grade Agentic AI & multi-agent systems for enterprise clients
• Building autonomous agent workflows across Manufacturing & Financial Services
• Driving applied R&D in agent accuracy, autonomy, scaling & efficiency
• Implementing enterprise AI guardrails & Responsible AI governance

Role & Experience

• AI Architect, Applied R&D & Industry Solutions @ Suzega
• 10+ years in AI/ML Engineering
• ex-EPAM | ex-Accenture | ex-TCS
• Location: India (Remote)

Key Expertise Areas

Multi-Agent Systems & Agentic AI LangGraph, CrewAI, OpenAI Agents SDK, AWS Strands Agents, AutoGen

LLMs & Fine-tuning PEFT/LoRA, Prompt Engineering, Model Guardrails, Agentic RAG

Agent Protocols & Interoperability MCP, ACP, A2A

NLP & Search NER, Semantic Search, Embeddings, RAG Pipelines

MLOps Real-time Inference, Latency Optimization, Model Observability

diff-guard  

Blast radius analyzer for AI-generated code changes. Detects unintended modifications by AI coding agents, measures downstream impact via import-chain analysis, and blocks dangerous commits before they reach production. Works as a git pre-commit hook with integrations for Claude Code, Cursor, Windsurf, and any AI agent. Supports Python (AST), JS/TS, Go, Rust, Java, and more.

Python CLI AST Analysis Git Hooks AI Safety PyPI

llm-finetuning

Reusable pipeline for fine-tuning SLMs on downstream tasks using PEFT/LoRA for efficient resource usage and faster inference.

Python PyTorch PEFT LoRA

GenNER

Named Entity Recognition using Generative AI approaches with Transformer models.

NER Transformers HuggingFace

med-text-classify

Medical text classification using NLP techniques for healthcare domain applications.

Jupyter NLP Healthcare

Emotion Classifier

NLP project for emotion detection in text using sequence classification and HuggingFace Transformers.

NLP Classification HuggingFace