* Merchant Credit Risk Tech Lead* Harness the power of LLMs and LLM agents to investigate risk and help sales, OPs, and strategy teams scale.* Maintain, improve, and launch models to assess risk, its magnitude, and opportunity cost.* Keep our code readable, reliable, and scalable.* Navigate the thrilling maze of migrating from one infra tool to next, and then to another, and, oh yeah, to another.* I think about offline and online metrics, live dashboards, rollout criteria, backtest results, interactions between multiple systems, quick evaluation of new features, and more.

- Team lead for ads measurement modeling (advertiser growth and solutions group)- ML solutions to increase user offsite visibility

I work on Uber's machine learning platform (Michelangelo team) with a focus on the deep-learning applications. I contribute to Horovod and work on productionizing deep learning models in Uber.Some of the tools I have worked on:- Horovod: Open-source distributed deep learning (https://github.com/horovod/horovod)- Petastorm: Open-source data access library developed- Ludwig: A declarative machine learning framework

* Lead engineer on data profiler and synthetic data generation platform* Utilized deep-learning architectures such as Generative Adversarial Networks (GAN) and LSTMs for generating fully synthetic datasets, detecting sensitive content in datasets and replacing them with seemingly authentic fake replacements, and identifying key/column types and track the changes.* Contributed to design and development of the infrastructure for on-demand deep-learning model training, hyper-parameter tuning, and model management.* Sought new customers. Assessed and re-assessed customers’ needs weekly. Completed review with C1 Risk Office* Reduced waiting time for accessing sensitive datasets in QA environments from days to hours, for 30+ teams and 400+ engineers* Filed 30+ patents on the developed technologies* Winner of Capital One Cape Award for Tech Excellence

Due to the growing performance requirements, embedded systems are increasingly more complex. Meanwhile, they are also expected to be reliable. Guaranteeing reliability on complex systems is very challenging. Consequently, there is a substantial need for designs that enable the use of unverified components such as operating systems without requiring their correctness to guarantee the safety of physical systems.My research is focused on designing software architectures that enable such a safety property for cyber-physical systems. In particular, I have proposed and implemented a novel architecture for fault-tolerance and fault-recovery that enables runtime system restarts with formal guarantees about safety during and after the restart. In addition, I proposed and implemented another novel monitoring kernel to utilize the combination of physical plant dynamics and ARM TrustZone technology to provide guaranteed safety.

* I was the 11th engineer at the time.* Contributed to Affirm’s Oauth 2.0 server, implemented the authentication of internal apps using Oauth* Developed an internal user dispute resolution and evidence management dashboard for customer service agents* Pushed code into production

* Implemented additional instrumentation of iTunes and AppStore applications on iPhone/iPad* Developed job-cycle for extracting consequent user activity sessions from unorganized event reports using Hadoop* Identified bottlenecks in content purchasing process and proposed improvements• Tested and submitted code into production.

Trained a malicious domain name detection using Neustar DNS report and deployed as a micro-service