Hemani Shah
@hemanishah
Data Scientist at ARThink AI
Valsad, Gujarat
Hemani is an experienced Data Scientist skilled in developing end-to-end AI pipelines. Expertise includes implementing multilingual speech-to-text systems, fine-tuning Vision Language Models, and creating POCs for facial recognition and customer service chatbots. Proficient in Python, Deep Learning frameworks, and various computer vision techniques.
Experience
Data Scientist
ARThink AI
Implemented an end-to-end pipeline for a virtual assistant that went through speech-to-text, RAG and text-to-speech. Implemented a pipeline for multilingual speech-to-text in Indian languages and translation into another language. Set up a pipeline to finetune Vision Language Model on custom datasets. Used various techniques like quantization and LORA for optimization. Created a POC for attendance management using facial recognition with video-to-video comparison, to capture different facial expressions, lighting conditions, and angles for each user. Created a chatbot for customer service at an airport, explored structured agents and OpenAI function calling to call various functions that gave information related to various services offered by the airport like parking, bus services, etc. Used various techniques like prompt engineering and RAGs to ensure accurate information was given. Developed an automated image processing pipeline for classifying spectacle prescription images and extracting key information. Implemented ROI detection using YOLO to standardize prescription information extraction. Established quality control measures, to ensure the accuracy and reliability of extracted data.
Data Science Trainee
DeetsDigital
Implemented Super Resolution to enhance the quality of images taken by drones to perform object detection. Automated the process of extracting information from invoices (in PDF format) and converted to Excel format the way the client needed. Used OCR to extract information from forms containing multiple languages. Created POC for sentiment analysis on human facial data using CCTV footage of people waiting in queues. Extracted information from resumes using OCR to map them to the most compatible job descriptions using LLMs to map them dynamically.
Teaching Assistant
Univ.AI
Worked under Faculty from Harvard University to develop coursework for Regression and Classification problems, Model selection techniques, Hypothesis testing, etc. Helped students debug code and understand ML concepts.
Quality Analyst Trainee Intern
Advanced Business & Healthcare Solutions
Created test cases. Learned various tools like Postman, Cypress, Zephyr and Jira.
Education
Gujarat Technological University
Bachelor of Computer Engineering
Licenses & Certifications
Google Data Analytics Professional Certificate
Google IT Automation with Python
Univ.ai Master ML & AI
Univ.ai
Career Essentials in Generative AI
Microsoft and LinkedIn