Hiteshwar Singh

> First engineer, Led AI product development, from prototyping to implementation of Compound AI System driving innovative solutions and feature integration. > Developed advanced LLM methodologies for dense search and retrieval, hybrid RAG, document analysis, and function calling, along with summarizing and data extractions from multimodal documents. > Designed and implemented AI Agents, by building architecture for multi-agent collaboration. > LLM evaluation frameworks, combining custom metrics, LLM-as-judge, and synthetic data generation for evaluating models. > Built AI System to generated research-driven reports by integrating multi-modal data, extracting insights and via IDP, domain-reasoning prompts etc.

• Leveraged Large Language Models (LLMs) to extract insights from unstructured medical notes for clinical applications. • Processed Electronic Fetal Monitoring (EFM) data, anonymized sensitive information, and utilized PostgreSQL and Python for dataset creation. • Developed text2sql engine for clinicians to access patient data using Language Models. • Applied predictive analytics to improve EFM data interpretation.

◦Examining the co-occurrence of Chronic Kidney Disease within the National COVID Cohort Collaborative (N3C) dataset for Y2 Q3. ◦Conducting an inquiry into the correlation between acute COVID-19 cases and the advancement of pre-existing chronic kidney disease (CKD) in adults. Investigating both COVID-positive and COVID-negative cohorts. ◦Employed Pyspark and SQL to transform and analyze a dataset of 16 million patient records. Utilized these transformations for two purposes: - comparing post-COVID CKD progression against a composite benchmark, - and assessing changes in eGFR values before and after COVID infection.

◦Significantly enhanced process analyst bot capabilities through focused testing and refining of Language Learning Models (LLMs), ensuring improved performance. ◦ Thoroughly curated several versions of domain and task specific dataset by deeply investigation and researching on product specs and targets. ◦Conducted in-depth exploration of diverse Language Models, identifying their potential for our process analyst bot. Trained and fine-tuned SLMs (BERT, TAPEX, TAPAS) and LLMs (LLAMA models) using Pytorch-Lightning. ◦Achieved a remarkable ~30% accuracy boost by strategically applying domain adaptation and masked learning techniques enhancing the model(s) performance.

◦Taught basics of Artificial Intelligence, Machine Learning and Python programming language. ◦ Provided guidance and oversight on hands-on projects. ◦ Assisted in creating course plans, material prep and implemented them to ensure quality learning for students

◦ Collaborated with team of 6 people for developing a machine learning based detection system. ◦ Achieved reduction in training time of the model, by incorporating different feature extraction techniques on data. ◦ Implemented a variety of machine learning algorithms to evaluate the performance of the model on processed data

◦ Reviewed software requirements and designed test case scenarios to perform testing on Android and iOS Mobile applications. ◦ Prepared reports on tests carried out for development team to resolve bugs and check the usability of the software. ◦ Participated in design reviews and provided input on required product design, and potential problems

Spring 2023: CSE545: Big Data Analytics CSE544: Probability and Statistics CSE532: Theory of Database Systems CSE523: Advance Project (under Prof I.V Ramakrishnan) Fall 2022: CSE538: Natural Language Processing CSE537: Artificial Intelligence CSE527: Introduction to Computer Vision

Jammu, J&K