- Designed and implemented key features for the Celonis Process Copilot which is a conversational agent that allows our users to analyze their processes using natural language. Responsibilities include being the owner of the persistence layer to support RAG based architecture, re-designing our API, and generating PQL (a SQL relative) using LLMs to answer user queries. - Designed an implemented the Knowledge Layer API from scratch in a small team of 2 other engineers. This API provides customers with out of the box content for proprietary Celonis process knowledge used to analyze a business’s processes along with allowing customers to be able to customize Celonis knowledge and import their own business knowledge. - Refactored the Duplicate Checker App, Celonis' most widely used app, to have an improved runtime complexity from O(N^2) to O(NlogN). Empirically, this translated to the app being ~7x faster while using an order of magnitude of less memory. This led to being awarded the CeloAI FY23Q4 Top Achievement Award.

Doing research in the Peterchev Lab studying the dynamics of neurons exposed to uniform and FEM generated E-fields to simulate the effects of tDCS on the brain. Currently developing deep learning software that would be able to estimate neural response of neurons exposed to a certain E-field. Skills gained: MATLAB, SimNIBS, Tensorflow