Projects

AI & Machine Learning Projects

Cadence Smart Health-Focused Dog-Collar

Developed a multi-modal AI model fusing acoustic, movement, and biometric data to detect pet stress, fatigue and health anomalies. Built a deep learning-based bark classification model for emotion detection, implemented anomaly detection for vital sign/activity monitoring and designed a dashboard for delivering real-time personalized insights to pet owners and veterinarians.

Duration: Jan 2025 – May 2025 | Role: Founding ML Engineer

PyTorch Deep Learning Multi-modal AI Anomaly Detection

Movie Recommendation System (ML in Production)

Engineered a production-grade movie recommender with PostgreSQL, Kafka, and modular pipelines, enabling real-time evaluation and scalable data management. Implemented canary testing, CI/CD, and MLflow provenance tracking, ensuring reliable deployments and secure experimentation in a cloud-like environment.

Duration: Jan 2025 – May 2025

PostgreSQL Kafka MLflow CI/CD

Research & NLP Projects

Retrieval-Augmented Generation (RAG) System

Implemented an end-to-end RAG using LangChain and Llama 2 to answer factual and event-based questions about Pittsburgh and CMU. Built a custom knowledge base, implemented document retrieval and optimized response generation for better accuracy.

Duration: Jan 2025 – Mar 2025

LangChain Llama 2 RAG NLP

Enhanced LLM Explanation Quality for Mental Health

Fine-tuned MentaLLaMA-7B with explanation-aware and emotionally instructed training, improving explanation precision (↑BARTScore +0.4) via LoRA, teacher forcing, reward scaling, and controlled data augmentation.

Duration: Feb 2025

LLaMA LoRA Fine-tuning Mental Health AI

Neural Machine Translation using RNNs & Attention

Developed & evaluated 8 NMT models using LSTM, GRU, and attention mechanisms for Western-Indian languages, enhancing translation precision. Results published in IEEE Xplore.

Duration: Jan 2022 – Aug 2022

LSTM GRU Attention Mechanisms IEEE Publication

Let's build something together

Contact me