SRIKANT HIREMATH

Engineered and fine-tuned 15+ BERT transformer NER models for Trial Master File and SMPC (Summary of Product Characteristics) documents, saving 500+ FTE hours annually by automating the classification and NER tasks. Proficient in Prompt engineering techniques for LLM's (GPT, GPT-3.5, mistral 7b). Engineered prompts for LLM and deployed a chat application employing gpt-3.5-turbo-16k and LangChain library to discern eligible patients for a genuine clinical trial based on provided patient records and study information. This tool is designed for use by medical professionals. As a result, this initiative secured $700,000 in funding for the organization. Engineered prompts for LLM and deployed 3 chat applications for conducting question and answer sessions, leveraging GPT-3.5, the Langchain library, and Pinecone vector database. This application is capable of recognizing user tasks from their input text and selecting appropriate tools to accomplish those tasks, using LLM and the Langchain library, with contextual information drawn from the chat history. Architected high-performance APIs for chat applications. Orchestrated Azure Storage, services bus topic triggered Azure functions for serverless solutions, Azure AppInsights for logging and Azure Container Apps to deploy API docker image. Proficient in writing unit test cases and testing API with postman.

Data visualization, data cleaning, feature engineering, feature selection. Text preprocessing, text summarization, entity extraction, text classification using spacy, genism. Basics of LLM - Bert. Table extraction from excel, classifying cells of excel whether a cell is within the table or not. This model helped the financial analyst team to write claims worth of 1.3 million dollars. NER using spacy, Bi-LSTM. Model deployment using flask framework.

Python, R. Data science toolkits, such as Numpy, Matplotlib, Pandas, SciPy, Scikit-learn, etc. Exploratory data analysis, predictive model building.