Usama Ahmad

Worked on the improvement of an AI assistant using GPT models from OpenAI and Google's speech-to-text technology to generate capability diagnoses based on patient symptoms. Conducted research and development on Transformer models and Large Language Models, exploring and fine-tuning models like BERT, Falcon-7b, and Llama2-7b on custom datasets. Specialized in doctor-patient exam encounters using OpenAI Whisper for transcription and GPT-4 for scientific notes for EHR systems. Developed AI Voice Assistant and Chatbot for patient scheduling using GPT-4 and RASA framework. Focused on building LLM trustworthiness with safety filters and guardrails, categorizing and validating datasets to ensure performance and security.

Worked on the improvement of an AI assistant undertaking, making use of GPT-4 Turbo, GPT-4 and GPT-3.5-Turbo from OpenAI and Google's speech-to-textual content generation to generate capability diagnoses primarily based totally on affected person symptoms. Conducted Research and Development (R&D) on Transformer fashions, Large Language Models (LLM), exploring Hugging Face repositories, developing and optimizing pipelines for neighborhood trying out of fashions, education and fine-tuning LLM fashions like BERT, Falcon-7b, Llama2-7b on custom datasets. Specialized in doctor-affected person exam encounters leveraging OpenAI's Whisper for transcription and GPT-4 for generating clinical notes for EHR systems. Developed AI voice assistants and chatbots for patient scheduling using GPT-4, also worked on RASA framework chatbots. Contributed to building LLM trustworthiness, safety filters, and guardrails by classifying and validating anthopic datasets and evaluating performance metrics based on

Worked on improvement of an AI assistant using GPT-4 Turbo, GPT-4, GPT-3.5 Turbo, and Google's speech-to-text to generate capability diagnoses based on patient symptoms. Conducted R&D on transformer models and large language models (LLMs), explored Hugging Face repositories, developed and optimized pipelines for local testing, training and fine-tuned models like BERT, Falcon-7b, Llama2-7b on custom datasets. Specialized in doctor-patient exam encounters leveraging OpenAI's Whisper and GPT-4 for electronic health record note generation. Developed AI voice assistants and chatbots for patient scheduling, also worked with RASA framework. Built LLM trustworthiness, safety filters, and guardrails by categorizing and validating a dataset of 40k questions, with extensive testing and evaluation based on research metrics.