Manipriyaa Ramesh

Computer Science Engineering student specializing in AI/ML with hands-on experience building AI/ML systems, backend services, RAG pipelines, and real-time computer vision applications. Strong foundation in OOP, data structures and algorithms, debugging, database systems, software testing, and performance optimization. Demonstrated ability to take ownership of technical problems, analyze failures systematically, and deliver solutions in fast-paced environments.

Product Intern — Learning & Enablement - Adrenalin eSystems Limited

(2026-06 - 2026-07)

• Built and tuned a Retrieval-Augmented Generation (RAG) pipeline for an AI-powered talent intelligence system, integrating document processing, semantic chunking, embeddings, vector retrieval, and LLM-based response generation.
• Designed and evaluated retrieval workflows to improve answer relevance and identify retrieval failures, hallucination patterns, and missing-context scenarios through systematic testing.
• Prepared and annotated benchmark datasets to evaluate retrieval quality, factual consistency, and end-to-end system performance.
• Contributed to backend and API integration across retrieval, generation, and serving components of the system.
• Used AI-assisted development tools to accelerate debugging, code analysis, and multi-file refactoring while manually validating generated implementations for correctness and reliability.

Research Intern — Computer Vision & Machine Learning - National Institute of Technology Tiruchirappalli

(2026-05 - 2026-06)

• Built a real-time driver cognitive-state monitoring system capable of classifying seven emotional and attention states at approximately 30 FPS on CPU-only hardware.
• Designed a hybrid inference architecture combining personalized LSTM models, general sequence modeling, XGBoost classification, and rule-based decision logic.
• Processed 15,000+ samples across five public datasets and generated synthetic data to address class imbalance and improve representation of rare states.
• Identified and eliminated data leakage caused by overlapping sliding windows, non-independent folds, and preprocessing contamination by redesigning the evaluation pipeline using non-overlapping data splits, GroupKFold cross-validation, and held-out preprocessing.
• Developed a real-time facial feature extraction pipeline using OpenCV and MediaPipe FaceMesh for continuous inference.
• Performed systematic model evaluation using accuracy, precision, recall, F1-score, and cross-validation techniques.
• Co-authored a journal-ready research paper documenting the system architecture, experimental methodology, and evaluation results.

Senior Core Member — Technical Team - Ladies Hostel Students Committee, VIT

(2023-11)

• Part of a three-member technical team which later grew to a team of 16 members responsible for managing end-to-end technical requirements for large-scale campus events.
• Coordinated technical execution, troubleshooting, and issue resolution during major university events including Freshers' Day 2025 and 2026.
• Demonstrated ownership in coordinating tasks, resolving operational problems, and delivering under time-sensitive event environments.

Block Organizer - Vellore Institute of Technology

(2023)

• Managed residential block-level coordination, communication, event logistics, and volunteer coordination.
• Took ownership of planning and execution responsibilities while coordinating with multiple stakeholders.

Bachelor of Technology - Computer Science and Engineering (Artificial Intelligence and Machine Learning) - Vellore Institute of Technology (VIT) (2027)