Apr 2025 – Present · 1 yr 1 mo

Jan 2025 – Apr 2025 · 3 mos · Toronto, Ontario, Canada

• Built open source GPU profiling tools for ML workload optimization, focusing on performance analysis and bottleneck identification in training and inference pipelines.

May 2024 – Nov 2024 · 6 mos

• Developed retrieval-augmented generation (RAG) methodologies for explainable logical query answering over knowledge graphs under Prof. Ashton Anderson.

Jul 2024 – Present · 1 yr 10 mos · Toronto, Ontario, Canada

• I built and demoed stuff.

Jul 2024 – Aug 2024 · 1 mo · Toronto, Ontario, Canada

• Developed open-source GenAI Eval frameworks and chatbots for data visualization.

May 2024 – May 2025 · 1 yr · Toronto, Ontario, Canada

• Worked across DevOps and Infra. Increased parallel build capacity by 150%, reduced pod downtime across OpenShift clusters by 12%, accelerated monthly release cycle processing time by 97%, and reclaimed 1TB+ of storage through automated image cleanup.

Jan 2024 – Apr 2024 · 3 mos · Toronto, Ontario, Canada

• Selected to participate as a Research Assistant in PRISM (Preparation for Research through Immersion, Skills, and Mentorship).

Sep 2023 – May 2024 · 8 mos · Toronto, Ontario, Canada

• Improved existing methods for autocolorization of black and white images.
• Developed a model training pipeline using PyTorch for preprocessing, training, and evaluation.
• Created a model search pipeline to enhance the base model by incorporating additional channel
• multipliers, dropout, and Conv2D layers to optimize architecture selection.

Sep 2023 – Apr 2024 · 7 mos · Toronto, Ontario, Canada

• Engineered a novel knowledge source that extracts Knowledge Graph entities via NER from unstructured text using Large Language Models (LLMs) for a Natural Language Understanding (NLU) system.
• Preprocessed and tokenized text into BIO encoding, utilized PyTorch to finetune LLMs, and conducted a comparative study to determine the impact of specific entities on overall accuracy.
• Performed multi-objective hyperparameter optimization using Optuna, tracking metrics such as accuracy, loss, precision, recall, and AUC to gauge model performance.
• Containerized the model using Docker to facilitate inference and communication with other knowledge sources within the NLU system.

Sep 2023 – Dec 2023 · 3 mos · Toronto, Ontario, Canada

• Led a team of students to develop an open-source project with UNICEF to seamlessly clone Apache Superset dashboard within and across Superset instances, eliminating redundant building and testing of duplicate dashboard between staging and production environments.
• Reduced time taken to clone and rebuild dashboards across Superset instances ('environments') by 93.6% (from 15 minutes to 57 seconds) using a React frontend and a Flask backend to communicate with the Superset API.
• Designed a novel cloud-based version control system using Flask (Python) and Firebase to allow updates to a dashboard to be reversed and be propagated to all its clones.

Apr 2023 – May 2024 · 1 yr 1 mo

Jan 2023 – Aug 2023 · 7 mos · Toronto, Ontario, Canada

• Developed reusable, stylized components in React and TypeScript for the hackathon’s landing page, driving 3000+ unique views and 1200+ hacker applications.
• Utilized the Notion API to fetch and display hacker information, schedule, and resources to 250+ hackers

Jul 2022 – Aug 2022 · 1 mo · Gurugram, Haryana, India

• Developed an automated testing suite consisting of unit, integration, end-to-end, and usability tests for
• an IoT waste collection solution.
• Automated >97% of tests using WebdriverIO and Selenium, reducing testing time by 99.1% (from 10
• minutes to 32 seconds).
• Facilitated identification of failed tests and increased ease of test analysis by automatically generating
• color-coded reports with screenshots and timestamps.