Somasree Majumder

Full Stack Machine Learning Engineer | AI/ML Specialist

Download Resume
Somasree Majumder

About Me

Full Stack Machine Learning Engineer specializing in computer vision, deep learning, and scalable AI solutions.

I'm currently a Full Stack ML Engineer at DeepEdge, where I develop cutting-edge anti-UAV quadcopter tracking systems with advanced object detection and tracking algorithms. With expertise spanning from YOLOv8 to Kalman filters and infrared thermal optimization, I've achieved over 50% performance improvements in challenging defense applications.

Previously at Deloitte, I architected multilingual RAG pipelines, built recommendation systems achieving 90% accuracy, and deployed scalable AI solutions using FastAPI and Docker. My experience also includes contributing to major open-source ML frameworks like PyTorch Lightning, Hugging Face Transformers, KerasCV, and KerasNLP.

As a published researcher in Homomorphic Encryption and ML in Medicine, Top 10 finalist at Challang Hackathon, and LiFT Scholar 2022 (Top 500 worldwide), I combine academic rigor with practical engineering to solve real-world problems through AI/ML technologies.

Technical Skills

Programming Languages

Python
JavaScript
TypeScript
C++
HTML5
CSS3

AI/ML Frameworks

TensorFlow
PyTorch
JAX
Hugging Face
Scikit-learn
Keras
NumPy
PaddlePaddle

Backend & APIs

FastAPI
Django
Pydantic
LangChain
LangGraph
REST APIs
GraphQL

Vector & Databases

FAISS
ChromaDB
Pinecone
MySQL
MongoDB
PostgreSQL
SQLite
Redis

DevOps & Cloud

Docker
Kubernetes
Git
CI/CD
AWS

Professional Experience

Full Stack Machine Learning Engineer

DeepEdge

March 2025 – Present | Hyderabad, India

  • Developed end-to-end anti-UAV quadcopter tracking system for runaway detection under challenging infrared conditions, delivering critical defense capabilities
  • Built advanced tracking models using YOLOv8 with sophisticated object tracking algorithms including linear and non-linear Kalman filters, and infrared thermal optimization for enhanced performance
  • Enhanced the performance over 50% over YOLOv8s by introducing motion queue masks
  • Managed 2-3 simultaneous client projects while maintaining high service quality and meeting all project deadlines. Researched and implemented findings from industry research papers like ERFE to achieve measurable performance improvements
  • Introduced a new technique: MOG2 Background Subtraction + Morphological Cleaning → Contour Detection + Size/Shape Filtering + SVM Classification of Motion Candidates which achieved 30% performance gains over YOLOv8
  • Implemented cutting-edge point tracking algorithms including Lucas-Kanade tracking and TAPNet by GoogleDeepmind, CoTracker by Facebook for precise tracking of point-like objects in complex environments

Machine Learning Engineer and Developer

Deloitte

July 2023 – March 2025 | Hyderabad, India

  • Architected and deployed multilingual RAG pipeline with FAISS vector database management, FastAPI endpoints, and Docker containerization for scalable AI solutions
  • Engineered Django-based movie recommendation system utilizing advanced algorithms including content-based filtering, FunkSVD, and popularity-based recommenders for personalized user experiences
  • Developed high-performance predictive model for customer purchase likelihood achieving 90% accuracy through sophisticated feature engineering and ML techniques

Machine Learning Developer (Freelance)

Unify.ai

February 2023 – May 2023 | Remote

  • Implemented robust functions across TensorFlow, JAX, NumPy, PyTorch, and PaddlePaddle frameworks, increasing code coverage by 10% and enhancing system reliability
  • Led comprehensive unit testing with Hypothesis framework and managed array-api-test suite achieving 80% test coverage for production-ready code
  • Developed sophisticated Ivy OpConverters for TensorRT backend and mentored contributors through complex large-scale codebase navigation

Featured Projects

🤖 AI Agent Flight Scanner

Intelligent flight data analysis system using advanced RAG architecture

Tech: Groq, Ollama, DeepSeek-R1, LangChain, SQLite, ChromaDB, FastAPI

Built intelligent flight data analysis system using advanced RAG architecture with DeepSeek-R1-Distill-LLama-70B and LLama 3.2:3B models for comprehensive flight analytics.

View on GitHub

🛍️ Amazon Product Classification

Advanced multiclass classification using ALBERT transformer

Tech: PyTorch, Deep Learning, ALBERT

Engineered sophisticated feature engineering and preprocessing pipeline improving model accuracy by 10% using synthetic PyTorch datasets and ALBERT transformer for advanced multiclass classification.

View on GitHub

🍕 Django Food Delivery Platform

Comprehensive full-stack food delivery solution

Tech: Django REST Framework, SQLite, JWT

Architected comprehensive full-stack food delivery platform featuring vendor management system, proximity-based search, payment integration, dynamic menu management, automated tax calculations, real-time order tracking, and notification system.

View on GitHub

🧠 ChildMind Anxiety Prediction

Predictive healthcare model for early detection

Tech: TensorFlow, Gated RNNs, Deep Learning

Developed innovative predictive healthcare model using Gated RNNs to analyze children's physical activity patterns for early detection of problematic internet use, contributing to child mental health research.

View on GitHub

🔍 Custom Vector Database

High-performance vector database using HNSW graphs

Tech: Python, HNSW Graphs

Built high-performance lightweight vector database using Hierarchical Navigable Small World (HNSW) graphs for efficient similarity search and retrieval operations.

View on GitHub

Research Publications

IEMATICS 2020 - Homomorphic Encryption

Research on data compression techniques for secure transmission and storage using homomorphic encryption methods.

Read Publication

SPECTRUM 2020 - Application of Machine Learning in Medicine

Discussed how various deep learning models solve complex real-life medical problems and their practical applications.

Read Publication

Achievements & Recognition

  • 🏆 Challang Hackathon: Top 10 finalist among 100 elite teams shortlisted worldwide (2 participants per team) - demonstrating exceptional problem-solving capabilities on global stage
  • 🥈 HackerEarth ML Challenge: Ranked 18/2500 participants
  • 🥉 Cipla Competition: Ranked 12/1000+ participants in data science competition
  • 🌍 Google Code Jam for Women: Ranked 1.6k globally
  • 🎓 LiFT Scholar 2022: Top 500 candidates worldwide in developers category by Linux Foundation
  • 👥 GHC OSD'23 Mentor: Mentor for OpenSSF project at Grace Hopper Celebration Open Source Day
  • 🌐 Open Source Contributor: Active contributor to PyTorch Lightning, Hugging Face Transformers, KerasCV, KerasNLP - recognized for high-quality contributions to global developer community
  • 📜 Multiple Certifications: Deep Learning Specialization, NLP, CNN from Coursera

Education

Institute of Engineering & Management

Bachelor of Technology in Electronics and Communication Engineering

August 2019 – May 2023

CGPA: 8.94

Get In Touch

Let's Connect

seckroll16@gmail.com
7980872028
Hyderabad, India