AI INNOVATION CREATIVE CODING
About Me
I'm an AI Engineer who builds the full spectrum of intelligent systems, from deep learning models(PyTorch) to advanced Generative AI and Agentic AI agents (LangChain, LangGraph). I specialize in deploying these solutions as scalable, production-ready services using Docker and FastAPI.
projects
Smart Credit Orchestrator - AI Invoice-Collections Agent
A LangGraph-driven finance workflow for
automated invoice follow-ups, tone escalation, and audit logging. Built as an end-to-end
AI agent using an 8-node StateGraph with two
conditional edges and an automatic two-attempt retry loop for validation failures.
Developed a FastAPI backend with 6 REST
endpoints and a Streamlit dashboard spanning
five pages for invoice queuing, audit inspection, and live agent execution.
Implemented Pydantic schema validation,
prompt-injection detection, and PII masking, backed by immutable JSON audit
logs — shipping 20/20 passing tests across
escalation and validation modules.
Integrated Groq and LangChain for tone-specific
email generation across four escalation stages,
including automatic legal flagging for invoices overdue by more than 30 days.
Weighing Scale Detection System
An end-to-end computer-vision pipeline for
intelligent scale-display detection. Curated and annotated a custom dataset with Roboflow, applying extensive
data-augmentation techniques to improve model robustness and generalization.
Fine-tuned YOLOv8n, achieving 99.5% mAP@0.50, 99.17% precision, and 100% recall on
the validation set. Engineered a modular training, evaluation, and batch-inference
pipeline with structured JSON reporting, plus a primary scale-selection engine using
multi-factor scoring (area, centrality, confidence).
Shipped as an interactive Streamlit web
application supporting real-time detection and downloadable outputs.
AI Document Summarizer & Q&A System
A RAG-based document-intelligence system for
multi-document PDF summarization and Q&A. Built an end-to-end Retrieval-Augmented Generation pipeline using
LangChain, Pinecone (384-dimensional vector DB), and
Groq LLMs for fast, context-grounded inference.
Engineered a multi-document processing workflow
supporting batch PDF uploads with recursive text chunking (with overlap), semantic search
retrieval, and hallucination-minimized generation.
Deployed with a drag-and-drop interface,
real-time chat, and multiple summary modes (short, medium, detailed) — hosted on Vercel and HuggingFace Spaces.
Experience
& Timeline
AI Engineering Intern
LearnNex · Remote
- Developed two production-ready AI systems: a multi-domain stateful chatbot built with LangGraph, and a RAG-based document-intelligence system that summarizes and answers questions across multiple PDFs.
- Optimized inference pipelines, cutting average response latency by 68% (2.5s → 0.8s) via FastAPI backends deployed in Docker, powered by Groq LLMs.
Open Source Contributor
Hacktoberfest 2025
- Merged 6 PRs to repositories such as TheAlgorithms/Java and sktime, emphasizing testing and algorithm optimization.
Problem Solving — LeetCode
Competitive Programming
- Solved 800+ problems focused on data structures and algorithms.