1. Created an efficient filter-based graph representation learning framework with 4.4% accuracy gains over baselines on node classification tasks (Publication at NeurIPS ’23 and MLG-KDD ‘24).2. Developed a novel natural language to code (repo-level) generation method for LLMs, incorporating static analysis tools and leading to over 20% increase in test rate performance (Publication at DMLR-ICML ’24).3. Collaborated with a 5 member team to finetune an LLM within a day to maximize performance on accuracy, bias and fairness metrics across diverse benchmarks (awarded second place at NeurIPS ’23 LLM Efficiency Challenge).4. Engineered an efficient PyTorch based MetaLearning library, achieving 5x faster performance than existing libraries, by leveraging functorch to parallelize inner-loop updates.

1. Saved resource costs by optimizing allocation of compute and memory through React and Grafana dashboards.2. Enhanced operational efficiency by automating infrastructure maintenance, resulting in a savings of over 600 engineer hours through the development of Python/PowerShell scripts.

Under the supervision of Dr. Alexey Melnikov in the Moscow Institute of Physics and Technology (MIPT), as a part of my Undergraduate Thesis, worked on employing Graph Neural Networks to various problems in Quantum Computing in an attempt to speedup Quantum Algorithms. Researched the latest advancements in Graph Neural Networks and trained models on Quantum Data using different architectures.

As a part of the E2E QE team, worked on providing scalable solutionsfor the parallelization of Android End to End tests and for contextswitching between Android Tests and Web Tests. Had practicalexperience with concepts like CI/CD, Agile methodology, Test DrivenDevelopment and BDD. Worked on technologies like Spring Boot,Cucumber, Serenity BDD, Selenium, Appium, REST Assured and SauceLabs.

Created a first of its kind food ordering application for students on campus. Instant hit with around 50 percent of total revenue coming in through the app in the first two months. More than a thousand downloads, with 100+ orders per day on an average through the app. Ended operations as a result of the pandemic. Play store link- https://play.google.com/store/apps/details?id=com.MESSiahPackage.cmp&hl=en_IN&gl=US