Qasim Wani

I am a senior at Virginia Tech studying Electrical & Computer Engineering. Over the past year, I've been researching with Prof. Ming Jin (robust reinforcement learning) and Prof. Ruoxi Jia (graph representation learning).

I previously interned at Qualcomm, where I worked as a Machine Learning engineer researching interpretable graph neural networks for chip design recommendations.

I'm passionate about solving critical challenges in Machine Learning pertaining to distribution shifts and optimization. I'm particularly interested in leveraging underlying data and problem structure to build more robust generalizable models; deep learning models that utilize structural dependencies will be critical to achieving generalization.

Email / CV / Google Scholar / Github / LinkedIn / Twitter

News

Apr. '22	Excited to join the Kleiner Perkins fellowship program this summer!
Mar. '22	Excited to join the 8VC fellowship program this summer!
Feb. '22	Excited to be an ODF fellow.
Nov. '21	Our team won the CityLearn competition (BuildSys), beating next-best solution (out of 24 teams) by 120%.
Oct. '21	My project, LeetHub, has been used over 2 million times by 4K+ developers worldwide.

Publications

Zeroth-Order Implicit Reinforcement Learning for Sequential Decision Making in Distributed Control Systems
Vanshaj Khattar, Qasim Wani, Harshal Kaushik, Zhiyao Zhang, Ming Jin
Under Review, AISTATS 2021
project page / code

We propose a novel framework for adaptive descision making, combining classical optimization with reinforcement learning via implicit differentiation.

This paper supports our work for the CityLearn competition as part of the BuildSys conference.

Projects

	HitchHiqe (May. 2019 - Sept. 2020) Safe and convenient ride-sharing platform for college students. Press. Voted as a top Virginia Tech startup, HitchHiqe helps students find rides to/from University. In Nov. 2019, I developed Fraper, a Bayesian algorithm to automatically find the best rides from social media platforms, reducing search time by 530x. Fraper helped HitchHiqe grow from 8K to 20K+ miles under 2 months! In June 2020, we received a $20,000 acquisition for HitchHiqe. Partially fueled by the COVID-19 pandemic, in Sept. 2020, I stopped working on HitchHiqe to pursue challenging research problems in Machine Learning.
	Preventiv (Feb. 2020 - July. 2020) Intelligent hierarchical contact tracing app. Webpage. Developed one of the world's first contact tracing apps and made it available for open-source. Created a novel Graph based approach to contact tracing using Bluetooth Low Energy and differential privacy. Video. Engineering feat: Single-handedly developed the Android app in under five weeks. TraceTogether (Singapore) took 40 engineers two months! Aided Virginia Tech with their COVID-19 response to 40K+ users. Left the project after doing my part to curb the spread of coronavirus.
	LeetHub (Nov. 2020) Automatically sync Leetcode solutions to GitHub in under 400ms. Source. LeetHub chrome extension has been used over 8 million times, helping students and developers with their coding interview prep. An average of 15,000 public commits are made daily via LeetHub. Turned down three acquisition offers to keep the chrome extension open-source; Received over 1.5K stars on GitHub with 400+ forks. 82% weekly active users. $0 spent on advertising! Took 24 hours to launch: Idea → Code → Publish!

Updated April 2021.

Design and source code forked from Jon Barron's website.