Headline

AI Engineer | ML, AI Agents, System Design, Generative AI, MLOps

About

Passionate AI Engineer with experience in designing, developing, and deploying cutting-edge AI/ML solutions. My expertise lies in Generative AI, Large Language Models (LLMs), Natural Language Processing (NLP), and Computer Vision, with a strong focus on delivering scalable, efficient, and impactful AI systems. What I Do: Generative AI: Build end-to-end Generative AI pipelines, including prompt engineering, LLM orchestration, and integration with vector databases (Pinecone, Chroma, FAISS). AI Agents: Develop autonomous agents and multi-agent systems using frameworks like LangChain and AutoGPT for tasks such as workflow automation, decision-making, and intelligent task delegation. Cloud & MLOps: Deploy AI solutions on AWS, Azure, and GCP, and ensure seamless integration using Docker, Kubernetes, and MLOps practices for continuous monitoring and retraining. Technical Skills: Programming: Python, PySpark, JavaScript. AI/ML Frameworks: TensorFlow, PyTorch, Keras, Hugging Face Transformers. Generative AI: GPT, LLaMA, Mistral, Prompt Engineering, Fine-Tuning, Quantization. AI Agents: LangChain, Langgraph, AutoGPT, CrewAI. NLP: Tokenization, Text Classification, Named Entity Recognition (NER), Summarization. Cloud Platforms: AWS (SageMaker, Bedrock), Azure. DevOps & MLOps: Docker, Kubernetes, Git, CI/CD Pipelines. Tools & Libraries: NumPy, Pandas, SciPy, FastAPI, Airflow, Databricks.

Location

Hyderabad

AI stack

antigravity

codex

opencode

Work experience

Software Engineer at MetLife, Jul 2025 - Present · 1 yr 2 mos, Hyderabad, Telangana, India, Building production AI systems at MetLife focused on agentic AI, MLOps automation, and ML infrastructure for the digital insurance platform. 1. Lead Nurturing Multi-Agent System Architected a 10-agent orchestration system using LangGraph and GPT-5 to automate lead qualification, scoring, and nurturing. Designed two agentic flows — Lead Qualification (Data Collector → Intent Analyzer → Qualification Scorer → Action Recommender) and Lead Nurturing (Behaviour Tracker → Engagement Strategist → Messaging Personalizer → Channel Scheduler → Propensity Scorer) with a Dormant Revival sub-flow. Built on FastAPI and PostgreSQL with real-time XGBoost propensity scoring, RAG-powered multi-channel outreach (WhatsApp, Email, SMS, Push), and human-in-the-loop compliance gates. Projected conversion uplift from 7.1% to 12% (traffic-to-lead) and 13.6% to 20% (lead-to-sales). 2. Propensity-to-Pay Model — MLOps Pipeline Delivered a production MLOps suite spanning data ingestion to model inference, achieving 80% ROC-AUC. Built Azure DevOps CI/CD pipelines and Azure ML pipeline YAMLs with Optuna hyperparameter optimization, champion/challenger evaluation, and cross-subscription Docker image promotion with approval gates. Reduced deployment from days to a single pipeline run with MLflow tracking. 3. MLOps Automation Agent Built an MLOps agent using LangGraph with intent routing and MCP server integration — enabling ML teams to scan repos, generate CAF-compliant artifacts, and trigger commits via natural language. Delivered 100+ ML pipeline templates with LLM-powered validation and a React streaming chat UI. Tech: LangGraph, GPT-4o, FastAPI, Azure ML, Azure DevOps, Docker, XGBoost, MLflow, MCP, Optuna, React.

Technical Engineer at Mercedes-Benz Research and Development India, Sep 2020 - Jun 2025 · 4 yrs 10 mos, Bengaluru, Karnataka, India, Built AI-powered systems for the wiring diagram and service documentation teams at Mercedes-Benz R&D India, focusing on RAG chatbots, computer vision, and ML classification. 1. Service Manual RAG Chatbot Built a production-grade RAG chatbot over Mercedes-Benz service manuals containing mixed content — procedures, tables, torque specs, and images. Engineered a multi-modal ingestion pipeline using PyMuPDF and pdfplumber, preserving the natural document hierarchy as retrieval units. Implemented hybrid retrieval (Qdrant + OpenAI text-embedding-3-small + BM25) fused via Reciprocal Rank Fusion, with MMR diversity reranking and Cross-Encoder relevance scoring before GPT-4o generation. Evaluated using RAGAS (faithfulness, answer relevancy, context precision). 2. RAG Visual Table Extraction Chatbot Built a RAG chatbot enabling the wiring diagram team to query and extract tabular data from 150+ industrial SVG and TIFF documents via natural language. Used ColPali to embed pages as visual vectors — capturing table structure, layout, and colour-coded wiring that text parsers cannot extract from TIFF files. GPT-4o interprets visual context and generates structured CSV output. Reduced manual data entry by 60% and improved accuracy by 90%. 3. Harness Detection Classification Model Trained an XGBoost binary classifier to detect whether a wire component is a harness, using features like voltage, colour, gauge, and insulation type. Deployed as a FastAPI endpoint in the wiring design workflow for real-time BOM validation. 4. Unified Data Integration Interface Built a Streamlit + MongoDB dashboard consolidating wiring data from multiple APIs, reducing data retrieval time by 50%. Tech: ColPali, GPT-4o, Qdrant, BM25, LlamaIndex, XGBoost, FastAPI, Streamlit, RAGAS, MLflow.

Graduate Engineer Trainee at Mercedes-Benz Research and Development India, Aug 2019 - Sep 2020 · 1 yr 2 mos, Bengaluru, Karnataka, India

Assistant Manager at Volvo Trucks India, Jan 2019 - Jul 2019 · 7 mos, Bengaluru, Karnataka, India, Development of Aftermarket Extended products.

Education

Master of Science - MS, Artificial Intelligence and Machine Learning, Scaler

Bachelor's degree, Automobile/Automotive Mechanics Technology, Lovely Professional University