Portrait of Apurva Umredkar

Apurva Umredkar

AI Engineer | Research Enthusiast

AI Software Developer at Vaspian, building RAG chatbots, conversational AI agents and MCP integrations. MS in Computer Science & Engineering (AI/ML) from the University at Buffalo, with prior experience in SAP integrations at Capgemini and published research in edge computing.

Experience and Education

AI Software Developer

02/2026 – Present

AI Product Development Intern

09/2025 – 11/2025

Vaspian LLC

  • Gained practical experience in integrating MCP servers, deploying containers with Docker, and managing version control using Git.
  • Developed a RAG-powered FAQ chatbot using n8n workflows and Bluemesh AI Studio, leveraging prompt engineering, structured output generation, and a knowledge base built from call transcripts.
  • Created a conversational AI agent in Eleven Labs and integrated it with Airtable using webhooks to automate the management of webinar invitations and customer response tracking.
Vaspian LLC logo

Associate Consultant

04/2024 – 07/2024

Senior Analyst

08/2022 – 03/2024

Capgemini Technology Services India Limited

  • Integrated external systems with SAP BRIM/SOM, serving over 10,000 users and reducing order processing time by 40% through scalable REST APIs using SAP ABAP, ABAP CDS, and BRF+.
  • Collaborated with the data migration team to map over 500 fields across 5 production systems and led system integration tests for more than 25 interfaces alongside the security and QA teams.
  • Resolved an average of 8-10 ServiceNow tickets weekly, encompassing production bug analysis and feature implementation through technical investigation and code development in SAP ACM.
Capgemini Technology Services India Limited logo

Projects

UBAssist: RAG-powered University Chatbot

LLMs, RAG, Cursor, Hugging Face, Docker, FAISS, LoRA, QLoRA

An intelligent university assistant providing instant answers to student queries about courses, policies, and campus information.

View Project

Grayscale to RGB Image Colorization

CNN, UNet, Vision Transformers, GANs, Gradio

Transform black-and-white photos into vibrant color images using deep learning. Perfect for restoring old photographs or adding life to grayscale artwork.

View Project

Phase Classification in Sleep Signals using CRNNs

Deep Learning, Inception, GRU, EEG Signal Processing

Automated EEG sleep phase classification system that identifies Cyclic Alternating Patterns (CAP) in sleep recordings, eliminating the need for manual expert analysis.

View Project

Publication

Journal Paper

Improved search and rescue optimization based load balancing scheme of CCTV footage on edge computing environment

Proposed ISROLBS-ECE, a novel load balancing scheme for edge computing that utilizes a hybrid chaotic search and rescue optimization algorithm.

This research addresses the critical challenge of processing high-volume CCTV footage in real-time. By effectively distributing offloading tasks to neighboring edge servers, the proposed scheme significantly reduces latency and prevents server overload, outperforming existing methods in efficiency and responsiveness for critical applications like search and rescue operations.

Published In

Concurrency and Computation: Practice and Experience

Publisher

Wiley

Authors

  • P. Deshpande
  • S. Dinesh
  • A.G. Kothari
  • Apurva Umredkar

Technical Skills

Programming & Data

Python logo
Python
C++ logo
C++
SQL logo
SQL
NumPy logo
NumPy
Pandas logo
Pandas
PostgreSQL logo
PostgreSQL

Machine Learning & AI

PyTorch logo
PyTorch
TensorFlow logo
TensorFlow
Keras logo
Keras
Scikit-learn logo
Scikit-learn
Hugging Face logo
Hugging Face
Ollama logo
Ollama
n8n logo
n8n

Backend & DevOps

FastAPI logo
FastAPI
Docker logo
Docker
Git logo
Git
GitHub logo
GitHub

Let's Connect

I'm always open to discussing new research, AI innovations, or potential collaborations. Feel free to reach out!

LinkedInGitHubEmail
© 2026 Apurva Umredkar