Shreeya Bose
Results-focused AIML Intern with expertise in large language models, retrieval-augmented generation, and document intelligence systems. Developed production-grade RAG chatbots and DocFusion, an AI-powered document processing platform handling 500+ documents with 70% efficiency gains and 90% answer relevance. Proficient in Python, TensorFlow, PyTorch, and cloud deployment (AWS), with proven ability to design scalable architectures, optimize ML pipelines, and deliver measurable performance improvements across retrieval speed and system reliability.
Experience
AIML Intern
TrueZeniX Services Pvt. Ltd.
Feb 2026 – Present
- Engineered a Retrieval-Augmented Generation (RAG) chatbot leveraging LLMs and FAISS-based vector search to process 1,000+ documents, achieving 90% answer relevance and reducing retrieval time by 60%.
- Architected and deployed DocFusion, an AI-powered document intelligence system that extracts and summarizes multi-format files (PDF, DOCX, TXT), automating processing for 500+ documents while delivering a 70% efficiency gain.
- Architected production-grade systems with robust error handling and scalability improvements that reduced system failure rates by 30%.
Projects
Soil Fertility Prediction System
Engineered a machine learning model that analyzes 15 soil parameters to deliver 85% accuracy in fertility classification, enabling precise soil health assessment. Optimized the preprocessing pipeline through strategic feature scaling and Principal Component Analysis (PCA), achieving a 30% reduction in training time and accelerating model iteration cycles. Deployed the classification system to empower agronomists with actionable, data-driven insights for yield planning, resulting in a 40% improvement in crop selection efficiency and supporting higher-confidence agricultural decision-making across operations.
EventMate- Event Management Website
Architected and delivered a comprehensive MERN stack event management platform leveraging React, Node.js, and Express to enable seamless CRUD operations across 10+ concurrent events with fully persistent MongoDB data storage. Engineered a responsive React.js interface powered by Axios HTTP client, eliminating all full-page reloads through intelligent dynamic content updates, resulting in significantly enhanced user experience and application performance.
AES-CBC File Encryption System
Architected and deployed secure file transfer infrastructure leveraging AES-CBC encryption to guarantee data confidentiality across all transmission layers. Engineered and optimized the encryption/decryption pipeline, delivering 40% improvement in processing speed relative to baseline benchmark tests. Designed and implemented resource-efficient cryptographic operations achieving real-time performance compatibility on memory-constrained embedded systems while maintaining security standards.
PathoVision- AI Powered Breast Cancer Detection System
Engineered robust Flask APIs for real-time histopathology image processing and prediction delivery, achieving end-to-end inference within 10–15 seconds per image and enabling rapid clinical decision support. Architected and deployed a comprehensive ML pipeline using Python, TensorFlow, and Keras to preprocess, train, and rigorously evaluate deep learning models across 10,000+ histopathology image samples, with performance validation through accuracy metrics and confusion matrix analysis to ensure diagnostic reliability.
Education
SR University
B.Tech, Computer Science
Sept 2022 – May 2026
Skills
Certifications
- AWS Cloud Foundations
- AWS Academy Machine Learning Foundations
- CCNA: Introduction to Networks
- Theory of Computation (NPTEL)
Volunteer
Content and Creative Head
CSI Hyderabad Chapter
- Spearheaded content strategy and creative campaigns that enhanced member engagement and social media reach through innovative digital solutions.
- Led a team of 300 members across content creation, event promotions, and design workflows, driving coordination and execution of all deliverables.
Core Committee Member
Literary & Master Communicators Club
- Orchestrated 15+ campus events, workshops, and competitions annually, achieving 40% average student participation rates and 100% on-time delivery across all initiatives.
- Spearheaded cross-functional collaboration to streamline operations, optimize resource allocation, and improve overall committee efficiency.
Publications
- Hybrid BERT-Based Deep Learning Models for Sentiment Analysis: A Study on LSTM and GRU ConfigurationsIEEE Conference Paper - ICICI-2025June 2025
ISBN: 979-8-3315-3829-3