Koppisetty Vijaya Venkata Durga Avinash

About

Dual-degree student in Mechanical Engineering at IIT Kharagpur with experience in AI/ML, full-stack prototyping, and data-driven system development. Built projects spanning GraphRAG, recommendation systems, computer vision, and negotiation workflows.

Experience

• Summer Intern
  IIT Kharagpur
  Jun 2026 – Jul 2026

  Objective: Develop an AI-enabled precision-agriculture platform for farmer advisory and direct agricultural market access. Engineered AI listing workflows with Faster Whisper, LLM field extraction, and Groq Vision for produce-quality grading API. Developed an AI negotiation engine using regional prices, farmer asks, platform margins, FastAPI, WebSockets and React API. Implemented 3-round automated negotiation with Accept/Counter/Reject decisions for real-time farmer-buyer interactions.

Education

• 2028 II
  M.Tech. Dual Degree 5Y

  IIT Kharagpur

  Manufacturing Science and Engineering · 2023 – 2028

  Manufacturing Science and Engineering · 2023 – 2028
• 2023 NJ
  Qualification

  Narayana Junior College

  2023
• 2021 CT
  Qualification

  Sri Chaitanya Techno School

  2021

Skills

Tools / apps / platforms

Projects

• Advanced Evidence Graph RAG System
  LlamaParse, LangGraph, Neo4j, Qdrant, BGE reranker, Redis, Groq

  Built a GraphRAG system to ground technical-document reasoning in traceable evidence and reduce context losses. Parsed and hierarchically chunked files, integrating graph traversal with hybrid dense + BM25 retrieval and cross-encoder reranking; benchmarked recall and MRR against source-level evidence ground truth.

• A2ZKART-Autonomous AI E-Commerce Agent
  Llama-3, Groq, LangChain, Streamlit, PostgreSQL

  Developed an e-commerce AI agent to manage end-to-end shopping workflows. Created a full-stack platform with real-time wallet deductions, time-bound order cancellations, and secure administrative CRUD controls.

• AI-Powered Workout Intelligence
  TensorFlow, OpenCV, Python

  Built a real-time AI system that segments workout videos into repetitions and evaluates form quality. Processed 200+ videos, extracted pose features, and trained exercise-specific CNN-BiLSTM models.

• Neural Two-Tower Movie Recommendation & Ranking System
  TensorFlow, Python

  Built a personalized movie recommendation system using deep learning. Transformed ratings into interactions, trained a two-tower model, and evaluated top-K ranking metrics across all users.

Courses & certifications

• Applied Data Science Lab · WorldQuant University

⚽ Extracurricular activities

• Global Arts Society (GAS) Drawing Competition - 3rd Place

  Secured 3rd place in a state-level drawing competition conducted across the state as part of Global Arts Society.

• National Students Space Challenge (NSSC) '24 Data Analytics

  Developed a hazardous asteroid classification model and engineered orbital features for predictive accuracy.