---

Hi there, I'm Saundarya

3rd Year Computer Engineering • K.J. Somaiya College of Engineering
CGPA: 9.5/10 • Honors: 10/10 • Mumbai, India

---

About

name: Saundarya Subramaniam
role: AI/ML Research Associate
location: Mumbai, India
focus: 
  - Vision-Language Models
  - Reinforcement Learning  
  - Legal AI Applications
  - Human-in-the-Loop Systems

current_work:
  research: "Legal document analysis using LLMs & VLMs"
  internship: "Building Xelly - Excel trade analysis platform"
  learning: "Multi-Agent AI Systems & Advanced RL"

interests: [Open Source ML, AI Research, Hackathons, Accessibility Tech]

---

Experience

AGR Knowledge Services • AI/ML Intern • May 2025 - July 2025

Xelly - Intelligent Excel Analysis Platform

• Processes 30K+ rows with clustering algorithms using cosine similarity
• 50% faster preprocessing with Prophet forecasting integration
• Built with Python (Flask/FastAPI), React, Redis, Plotly visualizations

Somaiya ML Research Association • Research Associate • Aug 2024 - June 2025

Legal AI Research

• Researching legal document analysis using cutting-edge LLMs & VLMs
• Conducted ML workshops for 100+ students
• Focus on democratizing legal knowledge access

Software Development Center • Data Science Intern • May 2024 - July 2024

Hindi Poetry Generation

• Preprocessed 1000+ Hindi poetry samples for generative AI
• Trained LSTM models to mimic original poets' writing styles
• Advanced text normalization and tokenization pipelines

---

Tech Stack

Languages

AI/ML

Development

Databases

---

Featured Projects

Xelly AI based excel trade data clustering and analyser Flask + React.js WebApp Cosine Similarity clustering + DBSCAN Real-time data analysis Integrated forecast models Tech: React.js Prophet Python Flask RAG MCP	FASAL FORECAST Crop Disease Detection System Flutter + TensorFlow mobile/web app InceptionV3 transfer learning Real-time disease classification Environmental data integration Tech: Flutter TensorFlow Python Flask
Preference Bandits Biased Feedback Simulator Multi-armed bandit implementations E-Greedy & Thompson Sampling Bias analysis and fairness metrics Interactive Streamlit dashboard Tech: Python Streamlit NumPy Matplotlib	Quizzila Interactive Quiz Platform Full-stack web application User management system Real-time scoring engine Secure session handling Tech: PHP MySQL JavaScript Bootstrap

---

Research & Publications

"Comparative Analysis of Deep Learning Models for Crop Disease Detection"
Subramaniam, S., Majumdar, S., Nadar, S., Kulkarni, K.

• Evaluated 5 deep learning models on 130K+ crop images
• Achieved 95.76% accuracy with custom CNN architecture
• Demonstrated transfer learning effectiveness for agricultural AI

---

GitHub Stats

---

Achievements

Competitive Programming

Certifications
Reinforcement Learning - NPTEL • Multi-Agent AI Systems - IUCEE

Recognition
National Scholarship - Ministry of Culture • 4 Classical Music Examinations

---

Connect

---

Currently exploring: Vision-Language Models • Legal AI • Multi-Agent Systems
Open to: Research Collaborations • Open Source Contributions • AI Hackathons