Ujjwal Surajkumar Pardeshi

About Me

I'm Ujjwal Surajkumar Pardeshi, a Computer Science undergraduate with hands-on experience in AI/ML, deep learning, and Backend development. Skilled in developing production-ready ML systems, RAG-based applications, and computer vision solutions.

Currently pursuing B.Tech in Computer Science (IoT) at SRM Institute of Science & Technology (2022–2026) with a CGPA of 7.5. I've interned at Samsung Research Institute, published research at IEEE CONNECT 2025, and lead R&D at IEEE SRM.

7.5 CGPA

2 Publications

11+ Certifications

8+ Projects

Education

SRM Institute of Science and Technology

B.Tech in Computer Science with Specialization in IoT

Sept 2022 – May 2026 Chennai, India CGPA: 7.5

Experience

Research Intern

Jan 2025 – June 2025

Samsung Research Institute – Bangalore, IN

Conducted research on emoji-based sentiment analysis and engineered a custom dataset of 2,700+ emoji-sentiment pairs.
Leveraged BERT and Emoji2Vec embeddings for enhanced semantic representation and model performance.
Explored and benchmarked multiple models: Neural Networks, Random Forest, LightGBM, and ensemble techniques, achieving 86.18% accuracy on the custom dataset.
Submitted findings to IEEE CONNECT 2025.

BERT NLP Sentiment Analysis LightGBM

Undergraduate Researcher

Aug 2024 – Jan 2025

SRM – Advisor: Dr. Vaishnavi Moorthy, Chennai, IN

Developed the Smart Lift Occupancy Detection and Visualization System using OpenVINO models and Next.js frontend for real-time crowd visualization and informed elevator usage decisions.
Integrated Firebase Realtime Database for instant data syncing and monitoring.
Deployed Raspberry Pi as an edge server over campus WiFi for seamless operation.
Secured 80k rupees project funding; won "Best Innovative Solution" at AIOT Project Expo 2024.

OpenVINO Next.js Firebase Raspberry Pi IoT

Skills & Technologies

Languages

Python C++ C JavaScript Node.js SQL

Frameworks & Tools

FastAPI Flask React Next.js LangChain Postman JWT

AI / ML / DL

PyTorch TensorFlow Keras OpenCV Hugging Face YOLOv8 MiDaS OpenVINO

Cloud & DevOps

AWS GCP Docker Git CI/CD Linux Nvidia A100

Databases & Backend

Firebase Supabase ChromaDB Raspberry Pi WebSockets SendGrid

Projects

RAG-Based AI Chatbot

Context-aware AI chatbot with ChromaDB, real-time WebSocket messaging, automated lead capture and email via SendGrid, and persistent user data storage using Firebase Firestore.

FastAPI LangChain ChromaDB WebSockets SendGrid

Underwater Reflection Removal & Depth Estimation

Hybrid deep learning and sensor-fusion pipeline using YOLOv8, MiDaS, and OpenCV for underwater reflection removal, object detection, and depth-based localization. Presented at AI FOR OCEANS.

YOLOv8 MiDaS OpenCV PyTorch

AutoSense

Advanced vehicle and pedestrian detection system using DeepLabV3 and ResNet-50 with a 96% accuracy rate.

Python PyTorch Computer Vision

Fitmon

AI-powered fitness game using Godot 4.3, TensorFlow.js, and JavaScriptBridge for real-time webcam-based exercise tracking; supports multiple movements and runs natively in browser via Godot HTML5.

Godot 4.3 TensorFlow.js JavaScript

Aircraft/Drone Detection

Robust aerial object detection framework suitable for surveillance, aviation safety, and wildlife monitoring.

Deep Learning YOLO Detection

Exoplanet Detection

Used NASA's Kepler and TESS data to detect Earth-sized exoplanets using Lightkurve.

Astronomy Data Science NASA

Gesture based authentication system Screenshot

Gesture Authentication

Security system enabling access via specific user gestures using sensors and machine learning algorithms.

ML IoT Security

Smart Lift Occupancy Detection Screenshot

Smart Lift System

Optimizes elevator usage on campuses reducing wait times and improving infrastructure management.

IoT Optimization Smart Campus

Publications

Under Review

PINNTO – Physics Informed Neural Networks for Trading Options

Ujjwal Pardeshi, Mehul Ashra

B.Tech Major Project • Dec 2025

Research applying physics-informed neural networks to financial options trading, combining domain knowledge constraints with deep learning for improved modeling accuracy.

IEEE CONNECT 2025

Emojis as Emotional Markers: A Computational Approach to Sentiment Analysis

Ujjwal Pardeshi, Sharanya, Dhruv, Achintya, Dr. Vaishnavi Moorthy, Dr. Pushpalatha (SRM), Harichandana B.S.S, Sumit (SRI-B)

IEEE CONNECT, June 2025

Computational approach to analyzing sentiment through emoji usage patterns, leveraging BERT and Emoji2Vec embeddings on a custom 2,700+ emoji-sentiment dataset.

Leadership

Head of Research & Development

IEEE SRM – Chennai, India

April 2023 – June 2025

Submitted journal paper titled "Evaluating Diverse Machine Learning Techniques For Autism Spectrum Disorder Detection" – currently under review.
Lead a team of 30 students; took KT sessions on Neural Networks, Machine Learning, SSL.
Presented 4 research posters at the 2nd International Conference on Higher Education Institute's Challenges and Solutions for SDGs'23.

Convener [Board Member]

Astrophilia – Chennai, India

Sept 2022 – Sept 2024

Acted as the liaison between Astrophilia and college administration, advocating for club interests.
Conducted engaging Astronomy sessions to enhance knowledge among members.
Organized and led team meetings to maintain project momentum and foster collaboration.
Awarded by the Directors of SRMIST for exemplary leadership and contributions to the club.

Certifications

Oracle Cloud Infrastructure 2025 – Certified AI Foundations Associate

Oracle Fusion Cloud Applications ERP Process Essentials Certified – Rel 1

Supervised Machine Learning: Regression and Classification

AWS Academy Graduate – Machine Learning Foundations

Classify Images with TensorFlow on Google Cloud

Artificial Intelligence Fundamentals

Data Fundamentals

Attention Mechanism

Introduction to Generative AI

Introduction to Large Language Models

Data Transformation in the Cloud

Ask Me Anything!

Online — Powered by Fuse.js

Get In Touch

I'm always open to discussing new projects, creative ideas, or opportunities to be part of your vision.