Aviraj Newatia Email & Phone Number

Cambridge, Gb

Toronto, On, Ca

Toronto, Ontario, Canada

My focus in my position as Director of Academic Affairs is to create a richer academic and research environment for undergraduates.I have initiated, and I am leading, the creation of the first-ever STEM Undergraduate Research Journal at the University of Toronto, St. George Campus. The journal aims to provide a platform for undergraduate students to publish their research in Computer Science, Mathematics, Physics, or Chemistry with plans to expand to other STEM fields.Additionally, I have spearheaded the running of a biweekly seminar series called the CSSU Undergraduate Colloquium, where undergraduate students present their research to their peers. This initiative aims to foster a community of undergraduate researchers and provide a platform for students to practice their presentation skills.Furthermore, I am working towards running research poster sessions for computer science students who involve themselves in research through one of the many programs at the University of Toronto, such as the Research Opportunity Program (ROP), Work-Study program, through volunteering, or through the independent faculty supervised projects (CSC494/CSC495). The poster sessions aim to provide students with an opportunity to showcase their research to their peers and faculty members and gain valuable feedback and experience in presenting their work.In collaboration with multiple different organisations, I have played an instrumental role as part of the Computer Science Student Union in facilitating, organising, and running HackThe6ix 2024, the inaugural Toronto Bioinformatics Hackathon, and the upcoming EigenAI Conference.I have also been created a mentorship program for incoming students to the Computer Science program at the University of Toronto. The program aims to pair incoming students with upper-year students to help them navigate the program and provide them with a support system.

Toronto, Ontario, Canada

I am working as a Research Intern under Dr. Rahul Krishnan studying Machine Unlearning methods for data and class removal and forgetting. We are targeting data-free machine unlearning methods, and ways to understand optimizing across changing loss landscapes for optimal performance without resetting.

Toronto, Ontario, Canada

In this position I have been collaborating and working under Professor Rahul Krishnan on Machine Unlearning for class-wise and item removal. I have been focusing on studying methods that use knowledge distillation, and those that do not require access to the entirety of the data.

Toronto, Ontario, Canada

In this position I am a member of the RNALab at the Donnelly Centre at the University of Toronto working under Dr Artem Babaian where we investigate the intersection of artificial intelligence and biology. This work focuses on examining the use of large language model embeddings on large quantities of data from the NCBI Sequence Read Archive to extrapolate and automate correspondences to uncover new insights across genomic samples and the classification of viruses. This includes analysis on 800,000 BioProjects and over 37 million BioSamples to draw connections between novel viruses and their project metadata.The work aims to explore new methods of studying the descriptiveness of text embeddings in the bioinformatics space. This is through exploring methods of separation with embedding basis vectors, and designing data manipulation methods to control and tune the way embedding models behave.

Toronto, Ontario, Canada

I am working under Director David Rokeby at the BMO Lab for Creative Research in the Arts, Performance, Emerging Technologies & Artificial Intelligence. My work at the BMO Lab currently focuses on developing a new architecture for real time interactive Audio Diffusion, building off of the previous work undertaken at the lab for the VoiceScroll framework with Stable Diffusion. This work aims to make use of freeze maps to allow AudioDiffusion models, which are largely limited to generating known time quantas of audio, to have the capacity of inference time interaction and coherent generation.Working with Professor Rokeby, I explored the use of open source Large Language Models in a theatrical setting using fine-tuned variants of the Llama-2 models. These implementations were made interactive, with real time capacity to modify inference parameters on the fly (Repetition Penalty, Temperature, Context Size) and blend different models during inference while maintaining coherence. I designed and implemented a novel data sampling mechanism to fit into the HuggingFace transformers library to improve sample efficiency for fine-tuning large language models on sparse text corpus' of data.Developed real-time (within 2 seconds) visual art generation with CLIP and VQGAN from prompts to create an interactive canvas with continuously optimizing and shifting latents using Python and PyTorch.Implementing Stable Diffusion techniques to artistic image generation and explored a tradeoff between accuracy and creativity in art generation using Python and PyTorch.

Toronto, Ontario, Canada

Worked with Professor Miriam Diamond and Professor Ziqing Hong to stress test, debug, and improve the functionality of the data catalogue suite and server for Super Cryo Dark Matter Search (SuperCDMS) SNOLAB at the University of Toronto. I constructed JMeter test suites to load test the data catalogue server software to pinpoint client and server-side data issues and corruptions and rectified data catalogue pipeline and server inefficiencies to improve data transmission and retrieval speeds and reliability.

Toronto, Ontario, Canada

Worked as a research volunteer under Professor Steve Engels during the summer after my first year at university as an undergraduate. The research we were conducting was focused on ways to improve the quality of mathematics education for primary children by using video games as a learning tool.In this role I attended weekly meetings with the other undergraduate researchers and Professor Engels in which we discussed development progress on games and research into this field which we were implementing in our video games.I developed a game targeted at mobile platforms and utilised the idea of a "hyper-casual" game based on ideas from games like Crossy Road and Frogger. The goal of the game was for children to learn and practice arithmetic and order of operations keeping in mind a goal to maintain (eg. a sum which was a multiple of 3).At the end of the summer, I demoed this game to a collection of researchers from a lab at the OISE (Ontario Institute for Studies in Education).The game is available here: https://project-avi.itch.io/pogocross.

Delhi, India

I identified places of improvement in the client onboarding process. Instead of manually specifying the main conversation topics and requests from customers after filtering through live chat conversation histories, I automated the process by developing an unsupervised learning clustering algorithm to process, read through, and group the most common families of requests and communication from customers of companies to speed up the customisation of the chatbot for new clients. This was accomplished using Python and PyTorch and utilising the Pandas and scikit-learn python libraries.

Gurugram, Haryana, India

I worked for the Nike and Rookie India divisions to analyze the current technical infrastructure in all of their departments and suggest and/or develop tools to improve the functionality of those departments.In this, I created an item recogniser model for ERP integration to decrease theft and increase in-store efficiency by reducing manual labour in Python and Tensorflow. I also developed a tool to form product associations from large sales datasets to improve product recommendations on the online storefront and also to provide further information to the visual merchandising department to improve their in-store designing and product placement decision process.Furthermore, I automated the customer feedback process by developing a recurrent neural network tool to analyse and classify customer feedback to provide information and insight about potential improvements to increase customer satisfaction.

Bachelor Of Science - Bs, Mathematics And Computer Science, 3.76 (Cgpa)

Activities and Societies: Director of Academic Affairs - University of Toronto Computer Science Student Union (CSSU), University.

Ib, Hl Mathematics Analysis & Approaches, Physics, Computer Science - Sl Economics, English Lit, Spanish, Hl 777, Sl 666

Activities and Societies: MUN Executive Board and Delegate, Swimming Team, Round Square Publication and Magazine Team, Team GuardianMiddle.