May 2024 – August 2024 | Dubai, United Arab Emirates
Worked on the award-winning Gen AI application, Tax Genie 2.0, as part of Deloitte AI Institute.
Built and deployed a production-grade Retrieval-Augmented Generation (RAG) system using LangChain, LlamaIndex, and Azure Cloud Infrastructure.
Designed and implemented the production MongoDB database and vector database for large-scale retrieval.
Led model evaluation and selection using Azure AI Studio, incorporating RLHF with industry experts.
Performed infrastructure-wide cost analysis to optimize resource allocation and client expenses on Excel.
July 2022 – August 2022 | Abu Dhabi, United Arab Emirates
Visualized client data to provide meaningful insights through dashboards using Power BI, Tableau, and Python (NumPy, Pandas, and Matplotlib).
Analyzed company data to generate performance reports using Microsoft Excel and Visual Basic.
May 2021 – August 2021 | Abu Dhabi, United Arab Emirates
Detected and fixed systemic data corruptions and inaccuracies from clients’ raw data using Audit Command Language and Microsoft Excel.
Technologies: Python, PyTorch, HuggingFace Transformers, BART, DeBERTa, GPT-4
Developed a zero-shot classification model to identify appropriate points of contact in corporate settings, using a dataset inspired by employee training programs. Employed BART and DeBERTa models, refining prompts iteratively to enhance classification accuracy within complex corporate communication scenarios.
Technologies: Python, PyTorch, CatBoost, LightGBM, Bloomberg Terminal
Designed and implemented an ensemble model integrating gradient boosting algorithms and neural networks to predict 5-day stock returns, achieving an MAE of 7.43% while successfully identifying directional movement. Extracted and processed market data from Bloomberg Terminal, engineering financial features including price action, volatility metrics, volume profiles, and technical indicators to capture trading patterns.
Technologies: NextJS, Python, OpenStreetMap API, Embeddings
Collaborated on developing a community memory map for Minnesota, enabling residents to locate and engage with their communities. Integrated an AI-powered search feature utilizing embeddings to find the closest events based on user queries. Leveraged the OpenStreetMap API for accurate geolocation and mapping functionalities.