• Conducting in-depth research on Large Language Models (LLMs) to enhance the study of genomic data.• Developing Foundation Models aimed at predicting the pathogenicity of genes.• Performing detailed analysis of gene sequences to gain insights into genetic variants.• Introducing a new training approach for LLMs, designed to interpret genome data more effectively than current state-of-the-art methods.• Fine-tuning GPT-3.5 and other OpenAI models with Medical NLP data, and conducting benchmarks against existing models using the USMLE exam metrics.

• Led the effort to scrape a customer's website for service information, employing Beautiful Soup and Selenium. Automated the process for over 100 service pages in both English and Arabic, optimizing data collection efficiency.• Collaborated with a team to integrate the gathered data into ChromaDB, fueling a robust RAG (Retrieval Augmented Generation) system. Utilized ChatGPT as the Language Model (LLM) in the LangChain framework to enhance language understanding.• Implemented the Automatic Speech Recognition (ASR) component of the project, incorporating Whisper for both English and Arabic, along with Google Cloud ASR.• Executed the deployment of the project on Gradio and contributed to the user interface (UI) design, ensuring a seamless and user-friendly experience.• Spearheaded a project overhaul by implementing Llama v2 as the base Language Model (LLM) for English data, replacing ChatGPT. This strategic move not only optimized API call time but also significantly reduced the cost of inference, demonstrating a keen focus on efficiency and cost-effectiveness.

• Conducted in-depth analysis of claim documents on Azure Databricks, systematically identifying emerging risks within organizational claims.• Parallelized Extract, Transform, Load (ETL) operations on claim documents using PySpark on Databricks, optimizing processing efficiency.• Applied advanced Natural Language Processing (NLP) techniques by fine-tuning RoBERTa and Longformer models, utilizing HuggingFace and Haystack, to extract the cause of loss from claims.• Devised and implemented a robust evaluation procedure, employing metrics such as Rouge and Bert-score, to assess the relevancy of extracted causes of loss in relation to the intended outcomes of emerging risks.• Fine-tuned a Longformer encoder-decoder-based summarization model, enhancing the ability to distill claim documents into concise cause-of-loss summaries.• Leveraged SparkNLP to fine-tune models for extracting the cause of loss and summarizing claim documents, showcasing a comprehensive approach to information extraction.

• Spearheaded the development of a Machine Learning engine with an impressive ~80% performance on various metrics, specifically designed to detect and extract employee information from resume documents.• Collaborated seamlessly with other departments to streamline the collection and organization of employee resumes, along with related metadata within the organization.• Successfully trained and deployed classifiers and Question and Answer (QnA) systems, enhancing the ability to classify resumes and extract valuable information from them.• Innovated the use of ArcFace loss-based embeddings to project skills from employee resumes into a multi-dimensional space. This groundbreaking approach facilitated the clustering and identification of various personas and roles among employees.

• As the very first engineer on the team, pioneered the development of a Proof of Concept (POC) for chemical matching based on their attributes, demonstrating innovation from its inception.• Led the establishment of the backend infrastructure on AWS, shaping the foundation for a scalable and efficient chemical matching system.• Applied advanced techniques such as semantic similarity with ChemBERT to elevate the accuracy of matching chemical text attributes.• Implemented a sophisticated matching system that spans eight hierarchical levels of chemical categories, achieving granularity and comprehensiveness in the matching process.

• Conducted cutting-edge research in Computer Vision, Retail Vision, and Edge ML.• Crafted sophisticated deep learning models using Tensorflow 2.0, PyTorch, and Flux.jl.• Innovated a Genetic Algorithm-based AutoML ensembler, streamlining the optimization of deep learning models.• Delved into state-of-the-art (SOTA) face recognition research, surpassing performance benchmarks set by existing systems.• Completed the development of a Binary Weighted Networks-based framework from the ground up.