Satyam Singh
Machine Learning Researcher | Neural Networks & Applied AI
IIT Mandi - Computer Vision Lab
About Me
Hey, I’m a Machine Learning Researcher | Neural Networks & Applied AI
I am a machine learning researcher and engineer focused on building efficient neural network architectures for real-world applications. My work spans deep learning, computer vision, and natural language processing, with a strong emphasis on representation learning, scalability, and deployment-ready AI systems.
I work on neural network–driven intelligence systems that operate under complex data distributions, limited resources, and real-world constraints, translating theoretical ideas into robust, practical models that perform reliably beyond controlled settings.
My goal is to create AI systems that are efficient, interpretable, and usable at scale, bridging cutting-edge research with production-ready deployment.
Current Research
Currently, I am working as a Research Intern at the Computer Vision Lab, IIT Mandi, where I focus on small object detection in high-resolution remote sensing imagery. My research specifically targets objects below 32 x 32 pixels under severe background clutter and scale imbalance, incorporating innovative multi-scale feature fusion with global context modeling.
Research Focus
My research contributions include:
- YOLO Architecture Redesign: Incorporating customized feature enhancement, multi-scale feature fusion, and global context modeling
- Contextual Representation Learning: Multi-branch convolutional structures and receptive field expansion
- Feature Fusion Strategies: Mitigating semantic inconsistencies between shallow and deep feature representations
- Global Context Modeling: Capturing long-range spatial and cross-channel dependencies
- Deployment Optimization: Optimizing inference latency, memory footprint, and parameter efficiency
Technical Expertise
Programming Languages: Python, Java, C++, JavaScript
ML/AI Frameworks: PyTorch, TensorFlow, OpenCV, NumPy, Pandas, Scikit-learn
Tools & Technologies: Git, Docker, Linux, SQL, MongoDB, React, Node.js
Recent Achievements
🏆 Finalist - IIT Guwahati Techniche Tech-Expo 2025 (Top 30 out of 2400 teams)
🥇 Rank 76 - IIT Kharagpur Data Science Hackathon (Kshitij) 2025
🏅 First Prize - College Research Paper Competition
📄 Research Paper Under Review - “IoT for Sustainable Resource Management”
When I’m not immersed in research, I enjoy exploring new technologies, contributing to open-source projects, and staying updated with the latest developments in AI and computer vision.
Download my resumé .
- Neural Networks & Deep Learning Architecture Design
- Computer Vision & Image Processing
- Small Object Detection in High-Resolution Imagery
- Multi-scale Feature Fusion & Context Modeling
- Remote Sensing & Satellite Imagery Analysis
- Medical Image Analysis & Healthcare AI
- Real-time Systems & Edge Deployment
- Model Optimization & Inference Acceleration
Computer Science and Engineering
Technical Skills
Core competencies in machine learning and computer vision
Experience
- Researched deep learning–based computer vision methods for small-object detection in high-resolution remote sensing imagery.
- Designed and optimized attention-based architectures and implemented GPU-accelerated models using PyTorch and CUDA for efficient training and inference.
- Applied models to agricultural and environmental monitoring datasets and contributed experimental results toward ongoing research publications.
- Collaborated with technology, operations, and capability teams to support the migration of GNA operations to the Genesys platform, ensuring system reliability and data integrity.
- Assisted in validation, testing, and monitoring workflows impacting 1,200+ agents during an 1-month phased rollout, contributing to stable production operations.
Projects
Research and Development Projects
Brain Tumor Classification Using CNNs
Transfer learning with pretrained VGG16 and ResNet architectures for medical image analysis, achieving 92% validation accuracy.
Neural Architecture Search
Automated neural architecture search techniques for efficient model design with multi-objective optimization, achieving 30% latency reduction.
Smart Power Demand & Generation Prediction System
Multi-region deep-learning system for optimal power generation forecasting using environmental, temporal, and demand data.
Sign Language Recognition System
Real-time vision-based system for American Sign Language gesture recognition with optimized CNN architectures, achieving 95% classification accuracy.
VARDAx Connect — ML-Powered Web Application Firewall
Production-grade machine-learning–driven WAF for real-time threat detection, anomaly identification, and automated security response.
Fake News Detection with LSTM
LSTM-based sequence modeling system for fake news detection using contextual understanding and pretrained GloVe embeddings, achieving 89% F1-Score.
Achievements
Competition Achievements
- Finalist - IIT Guwahati Techniche Tech-Expo 2025 - Top 30 out of 2400 teams, achieving top 1.25% performance in prestigious national competition. (2025)
- Rank 76 - IIT Kharagpur Data Science Hackathon (Kshitij) 2025 - Among top performers in one of India’s most prestigious data science hackathons. (2025)
Research & Academic Excellence
- First Prize - College Research Paper Competition - Research paper: “Fault-Tolerant Task Scheduling for Cloud Computing” focusing on novel fault-tolerance mechanisms for cloud environments. (2024)
- Research Paper Under Review - “IoT for Sustainable Resource Management” focusing on IoT-enabled resource monitoring systems. (2024)
- Outstanding Academic Performance - Maintained exceptional CGPA of 8.98/10 throughout B.Tech program at NIST University. (2023-Present)
Research Recognition
- Research Intern at IIT Mandi - Selected for competitive research internship position at Computer Vision Lab, IIT Mandi for work on small object detection in remote sensing imagery. (2024-Present)
- Publication in Progress - Research contributions in computer vision and small object detection being prepared for publication in top-tier conferences.
Technical Excellence
- Neural Architecture Search Innovation - Achieved 30% latency reduction through novel automated neural architecture search techniques. (2024)
- High-Performance AI Systems - Developed multiple AI systems with exceptional performance: 95% accuracy in sign language recognition, 92% accuracy in brain tumor classification, 89% F1-score in fake news detection.
Blog
No blogs available right now - Coming soon!
Creating a multi-page Dash Application
Building and deploying a multipage Python webapp in a few simple steps.
7 Must-Read Books for Data Scientists in 2022
Technical and Non-Technical books that will help you a become better data scientist.
Creating Dynamic Length Forms in Django
Discussing a fix to create a Django form which can have a dynamic number of input fields.
5 New features in Python 3.11 that makes it the coolest new release in 2022
Discussing the new features and updates in Python 3.11 and how to install the 3.11 Alpha version.
Data Science for Social Good Summer Fellowship
A summer fellowship for people looking to make a positive change in society through data science
Automated Text Analysis using Streamlit
Efficient and Quick Text Analysis Tool built using Streamlit including Text Summarization, POS Tagging and Named Entity Recognition.
Creating Multipage applications using Streamlit (efficiently!)
Building your first Multipage Streamlit application and deploying it. The prerequisite is knowing the basics of Python and Streamlit.
Downside Risk Measures — Python Implementation
Implementing Semideviation, VaR and CVaR risk estimation strategies in Python. Downside risk is when the returns go lower than the buy price and how to estimate them.
Asset Risk Management Strategy — Maximum Drawdowns
One of the key factors involved in asset and portfolio management is accurately assessing the risk involved in your investment. Implementing a stock market risk analysis strategy in Python.
A Novel Approach to Feature Importance — Shapley Additive Explanations
Machine learning interpretability is a topic of growing importance in this field. Discussin an aspect of machine learning interpretability — feature importance and a novel approach called Shapley Additives.
Predicting Reddit Flairs using Machine Learning and Deploying the Model using Heroku — Part 2
An end-to-end machine learning project series. Text Analysis and Model Building.
Predicting Reddit Flairs using Machine Learning and Deploying the Model using Heroku — Part 1
An end-to-end machine learning project series. Problem Definition and Data Collection from Reddit.
Poker Hand Prediction
An iterative approach to predicting which hand will win the match. An unconventional solution to a conventional problem.
Publications
Contact
- satyamsinghpx@gmail.com
- +91 78910 66189
- Berhampur, Odisha
- Connect on LinkedIn
- GitHub Profile