Aditya Jadhav

Experience

• 

  Embibe

  Jul 2020 - Oct 2020

  Academic Assistant

  During the internship I worked on curating problems on various mathematical concepts such as probability, calculus, linear algebra and statistics

• 

  Omdena

  Mar 2021 - May 2021

  Machine Learning Intern

  • Worked on end-to-end development and deployment of an object detection model, encompassing data cleaning, an- notation, model training, and hyper-parameter tuning to achieve optimal performance on a dataset of 10,000 images.• Developed a YOLO v5 model to track individual chickens and analyze their feeding behaviors, undertaking testing and optimization to ensure high accuracy and reliable results in improving health conditions of livestock.

• 

  RhythmFlows Solutions Pvt. Ltd.

  Jan 2022 - Apr 2022

  Software Development Intern

  • Assisted in development of a high-performance e-commerce platform using AngularJS, JavaScript, and Django. Implemented meticulous code optimization techniques, reducing latency and increased conversion rate.• Built user-friendly and responsive UI, enhancing user experience & increased product discovery through dynamic search.

• 

  Aquavit Pharmaceuticals, Inc

  Jun 2023 - Aug 2023

  Software Engineer Intern

  • Assisted in development of interactive, responsive and user-friendly web application for Perfectionist 1 & 2 (proprietary product suite) leveraging modern frameworks like AngularJS, ExpressJS, JavaScript, HTML, and CSS.• Implemented responsive design principles and optimized for seamless cross-browser and cross-device compatibility. • Created documentation to assist customers in understanding application capabilities, customizations, and best practices.

• 

  University of Southern California

  Jul 2023 - Sept 2023

  Graduate Research Assistant

  • Simulated experiments incorporating distributed training and implemented A-LiBi: Attention with Linear biases which enables extrapolation on longer sequences to predict mutation in RNA sequences. Actively documented simulations.• Fine-tuned DNA-BERT 2.0 as the base model benchmark and also evaluated a custom model on GUE benchmark.

• 

  Artificial Intelligence Institute of South Carolina

  Jun 2024 - now

  Research Fellow

  1. Fine-tuned Command R+, Mixtral 8x7B and Gemma-2 9B on Squad v2 using LoRA adapters.2. Evaluated the vanilla model, fine-tuned model and the instruction tuned version of each model for catastrophic forgetting.