Hey, I’m Karthik Uppuluri.

true-profile-small.png

Currently

Coming soon…

Previously

2017 - 2018, consulting as a product analyst / PM at startups and larger tech cos. Most recently, I helped Salesforce bring to life an in-app search widget to help millions of their customers easily search Salesforce help content through their Salesforce org.

2016 - 2017, I worked at Sentio to reduce the cost to be productive. Smartphones are everywhere, but being productive on them is difficult, therefore creators have to invest in expensive computing equipment. Sentio makes an affordable laptop shell that transforms your smartphone into a laptop by combining the computation of your phone with and a laptop-like interface. I focused on growth (among other roles). 

Before that:

  • Developed Xnote: A tool to annotate and share web articles from your Android device.

  • Worked at Pindrop Security to detect phone scams. A common phone scam is when a person calls you impersonating your bank, or another reputable institution, and asks you for sensitive personal information. We collected phone complaints from sites where victims of scams submitted complaints, and then we processed them to classify into a type of scam (e.g. Wells Fargo scams, or Microsoft support scams). We could then detect the rising and falling of scams in real-time. We published a paper, and our scam scanner was featured on Techcrunch after it recognized one on Tinder.

  • Undergraduate Research in Reinforcement learning under Jon Scholz - Georgia Tech: I implemented algorithms to control PACMAN’s brain in an existing Pacman Simulator, to help him score as high as possible. A straightforward approach would be to calculate, on each turn of Pacman, for each of his moves, all of the possible scenarios of the game that could occur, and then pick the move that would take Pacman towards a high score. But as you can imagine, the Pacman game has too many potential outcomes for this approach to be computationally feasible. The algorithms I implemented were textbook reinforcement learning algorithms that found clever ways to figure out the best next step without playing out all of the future scenarios in Pacman’s head.