🚀 Senior AI/ML Engineer | LLM & Agentic AI Specialist | MLOps | Deep Learning | Computer Vision | Fine-Tuning | CI/CD

📍 Senior AI Researcher & Developer — 5+ years architecting production-grade AI systems across healthcare, EdTech, legal, finance, and sales intelligence domains. Currently leading 8 concurrent enterprise AI/ML engagements at Chetu Inc.

• 🎓 Senior AI/ML Engineer with deep expertise in Agentic AI, LLM/SLM Fine-Tuning, RAG Pipelines, and MLOps
• 🏆 Promoted twice in 18 months at Chetu Inc. — Team Member → Senior TM → Technical Team Lead
• 💻 Skilled in LangChain, LangGraph, Google ADK, A2A, MCP, Unsloth QLoRA, LoRA, vLLM, RAGAS, LangSmith, Ollama
• 🤖 Experienced in On-Prem LLM Deployment (LLaMA 3.2 70B, 94% GPT-4o parity), Multi-Agent Orchestration, and Context Engineering
• 🚀 Built production AI systems across healthcare, EdTech, legal, finance, recruitment, and equestrian domains
• 📈 Expert in MLOps — MLflow, Docker, DVC, AWS EKS, GitHub Actions, GitLab CI/CD
• 🛠 Passionate about intelligent system design, enterprise AI integration, and measurable AI outcomes
• 📚 Always learning & building scalable AI pipelines with real-world impact

• Built a GPT-4o + LangGraph multi-agent system integrated with Automatiq B2B API for intelligent ticket routing.
• Designed an intent-routing state machine with a regex JSON sanitizer achieving 98% routing accuracy across 320+ test cases including adversarial and E2E journeys.
• Stack: LangGraph, GPT-4o, Automatiq API, Python

• Developed a real-time AI tutoring system with Whisper STT, OpenAI TTS-1-HD, and Weaviate Hybrid RAG (BM25 + Vector search).
• Implemented grade-aware LLM-as-Judge guardrails with WebRTC sub-300ms latency.
• Deployed on Amazon EKS with multi-stage Docker builds.
• Stack: Whisper, OpenAI TTS, Weaviate, WebRTC, AWS EKS, Docker

• Architected a 3-node LangGraph pipeline: Supervisor → DB Extraction → Self-RAG Vault with dual-layer guardrails (keyword + LLM).
• Achieved 98% routing accuracy and 100% guardrail block rate across 500+ validation tests.
• Stack: LangGraph, Self-RAG, Guardrails AI, Python

• Built a 7-node LangGraph StateGraph with Qdrant hierarchical memory (7 streams, MD5 dedup) and a 4-layer retrieval cascade.
• Integrated an LLM quality judge (0.75 threshold), improving retry accuracy by 30%.
• Stack: LangGraph, Qdrant, GPT-4o, Python

• Developed an Advanced Self-RAG chatbot using LangChain, LangGraph, MLflow, and Django.
• Supports multi-modal query handling (text, images, tables, flowcharts) extracted from PDFs.
• Achieved 87% answer accuracy on 100 human-evaluated Q&A pairs; BGE-M3 embeddings (nDCG 0.91); RBAC with 100% enforcement.
• Integrated Web Scraping with Selenium for real-time referenced data.
• Stack: LangChain, LangGraph, BGE-M3, ChromaDB, Django, MLflow

• Eliminated vector DB overhead entirely using JSON/in-memory cache (PageIndex approach).
• Built a dual-LLM pipeline (GPT-4o Vision + GPT-4o Text) with Gmail/IMAP integration for automated invoice extraction.
• Stack: Flask, SQLite, LangChain, GPT-4o Vision, PageIndex

• Designed a LangGraph multi-agent PPTX generator: Master Architect → Quantitative Analyst (Matplotlib charts) → Image Node (DALL-E 3) → Layout Reconciliation.
• Implemented AABB collision detection and dynamic font-scaling for polished slide layouts.
• Stack: LangGraph, GPT-4o, DALL-E 3, python-pptx

• Built a SequentialAgent pipeline: FintechAnalyst (Gemini 2.5 Flash) → LogisticsExpert (LLaMA 3.3/Groq) using Agent-to-Agent (A2A) protocol.
• Developed a Kafka KRaft FinBERT trading pipeline with BUY/SELL rules and a WebSocket real-time dashboard.
• Stack: Google ADK, A2A, Gemini 2.5 Flash, LLaMA 3.3, Groq, Kafka, FinBERT, WebSocket

• Deployed LLaMA 3.2 70B with 4-bit NF4 (bitsandbytes + Flash Attention 2) on 80GB GPU; achieved 94% GPT-4o parity while eliminating API costs.
• Built vLLM/Ollama OpenAI-compatible wrapper with GitHub Actions CI/CD for hot-swappable model iterations.
• Fine-tuned SLM (Llama 3.2 3B) via Unsloth QLoRA — 60% GPU memory reduction, 2× training speed; improved nDCG 73%→82%, MRR 68→71.
• Stack: LLaMA 3.2 70B, vLLM, Ollama, Unsloth, QLoRA, bitsandbytes, Flash Attention 2, GitHub Actions

• Built a Fusion RAG system (Pinecone + FAISS/SerpAPI) with MMR and cross-encoder reranking (ms-marco-MiniLM).
• Reduced report generation time by 60% (270s → 100s) via asyncio.gather; successfully stress-tested for 50 concurrent users.
• Integrated Stripe Connect for multi-tenant billing.
• Stack: Pinecone, FAISS, MiniLM, Django, Celery, Stripe Connect

• Built Text-to-Image generation models with Stable Diffusion, Pix2Pix, and Transformer-based architectures.
• Spatio-temporal crowd forecasting using YOLOv8 + CNN-LSTM, Gaussian density maps (ShanghaiTech dataset).
• AI-driven Image Captioning using BLIP2 and vision-language models; real-time object detection with Detectron2.
• U-Net + DLNN brain tumor segmentation/classification web app; ResNet-V3 plant disease detection REST API.
• Stack: YOLO, Detectron2, Stable Diffusion, BLIP2, U-Net, ResNet, PyTorch, OpenCV

• AI-powered meeting analytics system using YOLO (face detection), WhisperX (transcription, WER <8%), T5 (summarization, 92% grammar correction), Llama 3.3 (context analysis).
• Implemented Emotion and Sentiment Detection; automated Report Generation via Google Cloud API.
• Achieved 99% task reliability across 100 pipeline runs using Django Celery + RabbitMQ.
• Stack: YOLO, WhisperX, T5, Llama 3.3, Django, Celery, RabbitMQ, Streamlit

• Developed comprehensive web scraping frameworks using Selenium, BeautifulSoup, and Scrapy.
• Created intelligent data collection pipelines with cleaning and transformation techniques.
• Implemented ethical scraping practices (robots.txt compliance).

• BERT Autoencoder for query-based data extraction (ontology + PCA + WaOA clustering).
• NL-to-SQL pipeline using Seq2SQL + PSO for natural language database querying.
• Neo4j intent classification for graph-based reasoning; improved DL model accuracy 80% → 90%.
• Stack: BERT, RoBERTa, T5, Seq2SQL, Neo4j, PyTorch, Scikit-learn

Category	Technologies
Languages
Agentic AI & Orchestration
LLMs & Generative AI
Fine-Tuning & Embeddings
RAG & Vector Stores
Computer Vision
ML & Deep Learning
Web Frameworks
Messaging & Streaming
Databases
Data Processing
MLOps & Deployment
Cloud

🎯 "Transforming Industries with Intelligent AI Solutions" 🚀