2x founder. Engineer. Researcher.
ABOUT
I was brought up in the Bay Area, where I first learned how to play cricket. For 6 years, I represented my academy and region in national and international tournaments before I decided to take it a step further and play professionally in India. For the next 7 years, I went from captaining my district to captaining and representing my state at a national level.
Some snippets from over the years:
I came back to the Bay Area to pursue higher education, started community college, and got my associate's degree in computer science, and transferred to UC Irvine, where I'm currently a senior. I focus on the intersection of artificial intelligence and systems engineering.
Currently, I'm building Orderly, the multimodal AI ordering layer for every restaurant on Earth.
WORK
The multimodal AI ordering layer for restaurants. Guests scan a QR code, speak their order, and it routes directly to the kitchen through the existing POS. Signed a marketplace and co-sell agreement with Shift4 in under 30 days. Currently live in Orange Square (Santa Clara), soon rolling out to Fremont and Cupertino locations. Landed all 73 locations for Vitality Bowls, as well as 20 other founder-led restaurants. Signed 52 restaurants for the B&C group. Paid pilot with Flynn Group, the largest restaurant franchisee in the US, launching across Applebee's. Shift4 pilots still in progress.
SD-WAN Team
Architected and built the initial version of the no-code AI Agent Builder for Enterprise Automation.
Built and sold an AI workflow automation tool for construction with my childhood best friend and current Orderly co-founder, Math Heramia. Six-figure exit in under six months.
Predicting the specific DNA attachment site an integrase targets from its protein sequence is a challenge because of the missing co evolutionary signals in current representations. I built an ML pipeline using ESM-2 and DNABERT to analyze 491K+ integrase–att site pairs and biological structure with UMAP and HDBSCAN, and evaluate predictive models.
Did some cool research here as part of a small team on algorithmically matching dogs by personality. Collaborated with cross-functional teams for data analysis and architected solutions for recommendation systems.
RECENT PROJECTS
education
check out this neural network implementation that solves xor classification, which was a problem that proved single layer models weren't sufficient.
- toggle inputs directly by clicking on the digits
- press train to watch the numbers update live
- press 'test all' to see results
- press 'reset' to erase everything