Ansh Mehta

Assisting Prof. Klaus Mueller for the project on analyzing the strengths of LLMs like ChatGPT, and Gemini for Missing Value Interpolation and Estimation

Assisted Prof. Ayush Kumar and Prof. Gang Luo, and Harvard Medical School in their project on Brand Logo Detection for the visually impaired. Curated a dataset of 3337 images through Google Maps for 14 different brands. Trained Tensorflow models (Yolov5, Yolov8, RetinaNet, MobileNet), and achieved superior performance (Yolov5 mAP@0.5: 84.6%). Implemented a prototype on edge devices, deploying the model in an Android environment

1) Achieved an accuracy of 72.35% while developing a ResNet based the Discriminator of a Generative Adversarial Network model used for AWS SageMaker on Medical images (Diabetes Retinopathy). 2) Engineered Deep Learning systems, including their frontend and database, and deploy them with the help of Streamlit, Bootstrap, AWS, Pytorch, and the company’s environment.

Website - House Helpers Developed a Website called House-Helpers which focuses on providing house-hold services and help like Salon, Massage, Housekeeping, Trainer, Electrical Appliances and Plumbing Repairs. It was developed using Django and Hosted on Heroku. I used Django, Python, HTML, CSS, Bootstrap for developing the website while used Heroku as a platform to host it.

Basic Text Analysis Developed a Python code for reading Word and PDF files in Python using Docx and PyPDF2 libraries and also found the words which occur repeatedly. Created a whole Word-Cloud based on the received text. developed a basic website using Flask and hosted the project using Heroku.

Built an interactive website for personality trait identification, based on the OCEAN model, using a question and answer system, where the tech stack used was Firebase, PHP, Bootstrap, JS, and HTML. Collected data from a group of 1000 students of the Engineering Department using the system and performed analysis on the same, where the personality traits of Openness and Extroversion were common among the Engineering students.

Website- Admission Process Developed a Website on Admission Process of our College (which was soon going to be converted into university). Used Django, bootstrap, HTML, SQL, CSS.

I was a part of the Software development team at Pred Omic Health, which aims are providing cost-effective and user-friendly healthcare information management services to doctors, patients, diagnostic centers, and more components of the whole medical industry. Following is the summary of my contributions: 1) Designed and Developed a dashboard for monitoring product usage of the healthcare information management system. (I actively assisted the team in developing a dashboard for tracking usage of the system by the users and doctors, and displaying different KPIs through which the sales and marketing team can further assist the doctors or get insights into various aspects.) 2) Employed AngularJS for frontend, Django for Backend, and MySQL for interfacing with the database to revamp the PDF generation module of the application which was vital for doctors during medical report generation and Prescription generation. Additionally, helped in the conceptualization and ideation of the Hospital inventory module. It was such a great experience to be a part of a small but very effective team. Pred - Omic is a startup which meant that I got to know how dedicatedly businessmen, entrepreneurs, and engineers work for making an idea transform into a profitable business. I also got to know that not just ML/Dl, but simple applications like storing, processing, and analyzing data can simplify a doctor's and patient's life. Overall it was a wonderful experience and a great opportunity to learn.

Improved an existing ML model that was used for Financial Reconciliation that was built using Python and MongoDB (Studio 3T). Helped in reducing the latency resulting in improved performance. Cleaned the model by reducing the lines of code from 3400 to 2960. Wrote unit and service level tests using Selenium for the services for the application leading to a decrease of 14 bug reports related to the UI.

Learnt different Python Libraries and used them in my different tasks. Developed an decision tree classifier, a supervised and unsupervised learning model, all using python. Github Project link: https://github.com/mehtanansh/TSF-Internship