I architect enterprise-grade Agentic AI applications, high-precision Retrieval-Augmented Generation (RAG) pipelines, and resource-optimized local LLM deployments.
- AI & LLM Orchestration: LangChain, LangGraph, Ollama, Hugging Face, Vector DBs (PGVector, Qdrant)
- Backend & Full-Stack UI: Python, FastAPI, Nuxt.js, Vue.js, Tailwind CSS, JavaScript, CI/CD, Docker
- Data Engineering: Automated Web Scraping, Structured Schema Extraction, Context Engineering
-
Local AI RAG Assistant - Designed and deployed a Local AI RAG Assistant for HR behavioral reporting using Llama 3.2 on CPU-only infrastructure (8 CPU cores, 8 GiB RAM), achieving response time between 12 to 25 seconds without GPU — replacing tools that cost $20,000–$80,000/year commercially. The engineered dual-function backend architecture securely serves two distinct user paths: anonymous incident reporting for employees and interactive workflow guidance for HR staff, deployed fully on Azure Server to meet AI Data Privacy and Compliance requirements.
-
Field Audit Agent - The Field Audit Agent is a four-agent pipeline built with Pydantic AI on Groq, orchestrated by a FastAPI backend with a vanilla HTML/CSS/JS frontend. Each agent has one narrow job, a strict structured output schema, and a system prompt encoding specific guardrails.
App Demo: AI Demo -
Agentic AI Copywriter - Agentic AI Copywriter app using LangChain to aggregate data from authentic market sources, process it via autonomous AI agents, and serve a seamless user interface for creating platform-specific marketing materials. Designed, built, and deployed an end-to-end Agentic AI application using LangChain to orchestrate autonomous agents with DuckDuckGo and Wikipedia search tools. Engineered the full-stack architecture to aggregate data from online sources, process it via autonomous AI agent, and serve it via seamless user interface for creating platform-specific marketing materials
-
Multilingual AI Voice Assistant Backend - Streamlined internal corporate ticketing workflows, driving a 17% reduction in call volume. Developed the system logic to ingest voice inputs in 8 different languages, automatically populate HR forms, and route them to HR for follow-up. This automation eliminated manual online form-filling by 11% within the first month of deployment. (Confidential project – code cannot be publicly shared.)
-
AI Contract Summariser - A document summarization tool built for procurement workflows. Designed to run in the contract lifecycle just before a contract is sent to stakeholders for signature — giving reviewers a fast, structured summary of key terms and obligations instead of requiring a full read-through under time pressure. Ingests contract/RFP documents, chunks and embeds them, retrieves relevant sections via ChromaDB, and generates structured summaries using a locally-hosted LLM — chosen to keep contract content (pricing, vendor terms, etc.) from leaving the local environment.
Let's Connect: LinkedIn Profile | AI Demo Lab