Rishabh Shah Email & Phone Number

San Francisco, Ca, Us

New Jersey, United States

New Jersey, United States

- Designing and implementing advanced Retrieval-Augmented Generation (RAG) models with open-source LLMs to handle and query a dataset of over 1M documents.- Using MongoDB for high-performance and scalable data management, optimizing storage and retrieval of extensive open-source data.- Utilizing BGE-M3 embeddings to enhance document understanding and response generation

- Analyzed data from remote sensors in Kenya, developed a data pipeline, and created calibration methods to ensure consistent sensor readings in the same environment.- Working towards investigating the impact of pollution on human health using Oura ring data, and analyzing factors such as SpO2, stress levels, and heart rate

Durham, North Carolina, United States

- Leading the development of the Chatbot module for the extension project of Nurse Hackathon- Working to integrate LLMs like LLama, Falcon, and Mistral into low-latency systems, aiming to enhance their effectiveness and optimize for resource-constrained applications.- Developed persona based chatbots by fine-tuning models via LoRa and QLoRa using axolotl

Bengaluru, Karnataka, India

- Contributed to platform deployment and reporting for the Global Azure Team, utilizing Machine Learning and Azure Cloud expertise. Specialized in IBM's Content Intelligence platform for document intelligence.- Developed a Table detection Model(Computer Vision) and Similarity Search for Automatic Document Content Validation & processing ~100K Documents leading to 50% cost savings ($200-300 K per year) in the Manual Audit. Used Python, BERT, and Camelot for the same. - Compared Semantic Search capabilities of Azure Cognitive Search and Kira to extract Clauses from Legal Contracts (Amendments, SOW and Master Contracts.)- Developed a module for document meta-data extraction and hierarchy generation of connecting documents - Enhanced IBM's Content Intelligence platform by developing document classification and custom key-value pair extraction modules, while optimizing end-to-end document processing.- Used Azure Form recognizer to modify CI to analyze invoices and extract Key-Value Pairs.- Deployed end-to-end Content Intelligence platform for clients on Azure Kubernetes Service using Terraform and Jenkins.- Participated in IBM Neuro-symbolic Data Scientist Collaboration for research in Neuro-symbolic AI where PyTorch was used extensively

Bangalore

- Made rule-based classification model, LSTM model, etc., for classification of different types of clauses- Proposed and built two clause segregation approaches: Combined Random Forests with concrete features and other one utilized computer vision techniques.Made rule-based classification model, LSTM model, etc., for classification of different types of clauses

Bengaluru, Karnataka, India

- Contributed to the development of a PDF document search module for an intelligent search platform, leveraging ElasticSearch and applying retrieval techniques such as TF-IDF and BM25.- Used Transformers like BERT, XLnet for passage rankings

Master Of Engineering - Meng, Artificial Intelligence

Relevant coursework: Sourcing Data for Analytics, Modeling Process & Algorithms, Optimization in Practise, Deep Learning Applications.

Bachelor Of Technology, Computer Science

Ix-Xii