Indroneel Roy profile picture

Hello, I'm

Indroneel Roy

ML Researcher | Computer Vision | Deep Learning

My LinkedIn profile My Github profile
Arrow icon

Get To Know More

About Me

Experience icon

Experience

3+ years
ML Research

Education icon

Education

B.Sc. (Bachelor’s) in Electrical & Electronic Engineering (EEE)

I'm a machine learning engineer specializing in computer vision and agentic AI. I transform research papers into working implementations, building everything from image segmentation models to intelligent chatbots. My expertise includes ResNet, DenseNet, EfficientNet, transformers, and hybrid CNN architectures for detection and segmentation tasks. I love optimizing models for better performance and taking them from experimentation to deployment. When I'm not coding, I'm playing chess or cooking—both keep my problem-solving skills sharp.

Arrow icon

Explore My

Experience

Core ML & Python

checkmark

Python

Experienced

checkmark

PyTorch

Experienced

checkmark

Scikit-Learn

Experienced

checkmark

Pandas

Experienced

checkmark

NumPy

Experienced

checkmark

SQL

Intermediate

Classical ML Models

checkmark

Linear Regression

Experienced

checkmark

SVM

Experienced

checkmark

Random Forest

Experienced

checkmark

XGBoost

Experienced

checkmark

Decision Tree

Experienced

Computer Vision

checkmark

CNN Models

Experienced

checkmark

Hybrid CNN

Experienced

checkmark

Image Segmentation

Experienced

checkmark

Object Detection

Experienced

checkmark

3D Reconstruction

Experienced

Generative AI

checkmark

LangChain

Experienced

checkmark

LangGraph

Intermediate

checkmark

Agentic AI Chatbot

Experienced

checkmark

Multimodal Agentic AI

Intermediate

Deployment & DevOps

checkmark

FastAPI

Experienced

checkmark

Streamlit

Experienced

checkmark

AWS

Intermediate

Problem Solving

checkmark

LeetCode

200+ Problems Solved

checkmark

Algorithm Design

Intermediate

checkmark

Data Structures

Intermediate

Arrow icon

Ongoing Work

Current Research

Exploring cutting-edge solutions in computer vision and predictive modeling

In Progress

HybridCNN-Polyp-Segmentation

A novel hybrid deep learning architecture combining ResNet-50 CNN encoder with Transformer blocks for accurate polyp segmentation in colonoscopy images. The model leverages cross-attention mechanisms to fuse local spatial features from CNN with global context from Transformers, achieving superior segmentation performance on medical imaging datasets.

Key Objectives

  • Design and implement a hybrid CNN-Transformer architecture integrating ResNet-50 encoder with multi-head self-attention mechanisms
  • Develop novel cross-attention fusion module to synergistically combine local CNN features with global Transformer representations
  • Leverage skip connections and hierarchical decoding for precise polyp boundary delineation across multiple scales
  • Reduce computational complexity while maintaining high accuracy

Technology Stack

PyTorch U-Net ResNet Attention Mechanisms Medical Imaging
Progress 85%
In Progress

Earthquake Magnitude Prediction using Machine Learning

Developing an advanced machine learning system for predicting earthquake magnitudes using seismic data and geophysical parameters. The research leverages deep learning models to analyze historical seismic patterns, geological features, and real-time sensor data to provide accurate magnitude predictions and early warning capabilities.

Key Objectives

  • Build a robust prediction model using ensemble methods and gradient boosting
  • Integrate multi-source seismic data including P-waves and S-waves
  • Develop real-time magnitude estimation algorithms
  • Improve prediction accuracy for early warning systems

Technology Stack

Scikit-learn LSTM Transformers Time Series Analysis Seismic Data Processing
Progress 30%
Arrow icon

Browse My Recent

Projects

HybridCNN-Polyp-Segmentation
Deep Learning

HybridCNN-Polyp-Segmentation

Novel hybrid CNN architecture combining multiple CNN models for enhanced feature extraction and accuracy

PyTorch CNN U-Net ResNet-50
LangGraph Chatbot
Generative AI

Agentflow-Chatbot Multi-Tool Conversational AI

Multi-Tool Conversational AI with advanced agent capabilities using LangGraph and LangChain

LangGraph LangChain Python
Wheat Detection
Computer Vision

Vision-Based Wheat Detection

Deep learning model for automated wheat detection and analysis using advanced computer vision techniques

PyTorch Python OpenCV YOLO
Cassava Disease Detection
Computer Vision

Cassava Disease Detection

CNN-based system for identifying and classifying diseases in cassava plants using image recognition

PyTorch CNN FastAPI Streamlit
Arrow icon

Get in Touch

Contact Me

Feel free to reach out for collaborations or just a friendly chat

Email icon

Email

indroneel1000@gmail.com
LinkedIn icon

LinkedIn

Connect with me
GitHub icon

GitHub

View my work