Leveraging AI and LLM on a new food app to build custom meal plans based on takeout• Running custom scrapers simultaneously on personal server farm using Python, Selenium, and GCP to scrape millions ofdishes from thousands of restaurants• Building the backend using Python, FastAPI, SQL, and Weviate VectorDB• Creating dish searches using VectorDB and search bar RAG powered by LLM for custom search queries• Implementing a chatbot for users to inquire about restaurants, dishes, opening hours, calorie count, ingredients, and more• Fine-tuning LLM with scraped data to create a RAG application that generates a custom meal plan of takeout and deliverydishes based on user preferences

Leading the AI, ML and Data Science initiatives at AdTech firm with Python, LLMs, VectorDB, LangChain, AWS, CUDA, and SQL• Advertisement Data Generation:Used GPT-4 and custom NLP models to curate a database of 7,000,000+ user content preferences based on age, gender, and location from 100+ sources enhancing targeted advertising• Advanced Podcast and URL Categorization:Trained and deployed an NLP pipeline with custom models and BERT to predict Podcast and URL categories• Innovative Podcast Similarity Search:Engineered a similarity search system for podcasts using BERT, VectorDB, GPT-4, and Llama2, boosting content discovery and accessibility• Multilingual Contextualization and Brand Safety:a. Developed a pipeline for transcriptions and categorizations in English, Spanish, and French using Bert, custom NLP models, WhisperAI, AWS S3, and SQLb. Implemented a feature for custom topics targeting using GPT-4 and WhisperAIc. Auto generate comprehensive Brand Safety reports using Llama2 and GPT-4, ensuring content adheres to IAB guidelinesd. Deployed on EC2 GPU instance and is monitored, upgraded, and trained periodically• Local Document Chat Bot, Summarizer and Translator:Created and deployed a local chat bot, summarizer, and translator using Bedrock, Llama2, LangChain and VectorDB for sensitive documents up to 100 pages• Detailed Information Generation:Leveraged Llama2, GPT-4, LangChain and VectorDB to autonomously generate detailed information from websites• Email Automation: Implemented an auto email chat bot using GPT-4, Llama2, VectorDB, and AWS, optimizing communication and improving response times• AI-Driven Bid Optimization:Trained and deployed AI models for bidding optimization, resulting in a revenue increase of over $1 million for the company• Detailed System Testing and Reporting:Automate Python unit and end-to-end tests, system diagnoses, and generate detailed reports for diverse teams and presentations to executives and management

Researcher, working with NLP, financial and traffic data to improve machine learning model training and accuracy• Created a Python simulator to emulate the travel patterns of thousands of vehicles within a city• Developed patent pending AI for next generation navigation based on traffic and cellular datao Scraped and cleaned over 32 million data points using DataBricks and PySpark for training AI modelso Trained AI models in Python using Keras, TensorFlow, PyTorch, and SkLearno Built and deployed the ML pipeline on EC2 instance. Pipeline included periodic automatic updates and re training of models• Built custom Federated Machine Learning models using TensorFlow and trained on Fintech, Financial, and Statistical data• Researched new strategies and models to improve Automatic Speech Recognition (ASR)• Implemented BERT models from HuggingFace to analyze millions of online product reviews

TA for the following courses:• Artificial Intelligence• Functional Programming• Concurrent Programming• Operating Systems• Data Structures

Worked as a ML and Backend Engineer for a new health tech platform, used by over a million customers globally• Created APIs and primary functions using .NET Core, C#, and MongoDB• Developed a Recurrent Neural Network to translate company’s products to several languages• Built Regression and Classification Machine Learning models on over 100,000 industry provided data to predict next steps,chances of approval and length of current stage for client’s products• Trained an NLP classifier model on 8 years of PHI data to categorize patient’s instant messages into different behaviouralcategories in real time• Improved user experience and engagement by integrating Login with Username and a User Points Awards System features• Created the company’s mobile app from scratch with Python and SQL on the backend, Flutter and Dart Lang on the frontend• Led a team of five to fully automate unit and system testing using Selenium and Java to reduce software testing from days to mere minutes

Worked on Upstream Extraction Retrofitting process• Answer and coordinate RFIs, RFRs and TDRs from various contractors• Design and aid in the construction of several expansion joints for 4” pipes carrying polymers• Studied and created datasheets for motors and pumps of various size and power• Gained experience working with power distributions systems, generators, transformers, andswitchgears• Worked with Engineering Manager to update the process of handling and implementing "LessonsLearned" into the daily activities of the project• Participated in walk downs of Electrical and Instrumentation equipment on site

Worked on high end research related to robotics surgery with the latest instruments such as the KUKA robot• Presented Conference paper on "A comparison of standard and telecentric lenses for use in 3D Digital Image Correlation strain measurements" at the combined AOA/COA meeting, Montreal 2014• Published a paper on “Measuring Strain using a 3D calibrated system”

Worked with LabVIEW for designing several electrical circuits• Official tester for NI myDAQ for second year electrical engineering students• Designed and built several Brett Board circuits with NI DAQmx• Worked with Solid Works, designing simple structures