Worked with various AWS customers as a member of AWS' Professional Services organization. Focused largely on projects in generative AI, including operationalizing large language models (LLMs) and multimodal models. Also worked various projects in the computer vision (CV) domain. Example projects:- Built a LLM-powered chatbot system with retrieval-augmented generation (RAG) to act as a natural language interface to a customer's sales documentation.- Built a semantic segmentation CV system to segment various objects in satellite imagery of residential real estate properties, used in a downstream system to valuate said properties.- Built a object detection CV system to detect defects in manufacturing line imagery of vials of drugs, improving F1 score to >0.95 and savings $1MUSD+ yearly for our customer.- Built a semantic search system for a customer's PDF documentation with various AI features built on top of the search result set; including clustering, sentiment analysis, and summarization via LLMs.- Built a semantic search system for a customer's job search platform, allowing for candidates to search for relevant job postings, and for hiring managers to search for relevant candidates based on candidate work experiences.

- Built models for active users of 100K+ Android and iPhone apps- Leveraged panel data for various modeling approaches- Build and optimized large-scale jobs for calculation and serving of estimates

- Machine learning & reinforcement learning research and development with application to real-world problems- Applied cutting-edge batch reinforcement learning algorithms to large datasets and deployed models to a variety of real-world test scenarios- Implementation of methods from the off-policy evaluation literature to validate models prior to deployment

- Studying machine and reinforcement learning methods for spiking neural networks.- Developing CPU- and GPU-enabled spiking neural network software geared towards machine learning applications.

- Explored learned representations of deep neural networks using activity-based correlation measure.- Studied layer-wise classification of image data.- Developed unsupervised Hebbian-learning layer on the output of layers with learned parameters.

I am responsible for programming software for motor and cognition tasks for "psychology of sleep" studies, as well as performing data processing and analysis on the data that result. I take care of hardware and software issues that cannot be solved by the core staff, including software upgrades, network troubleshooting, and general computer maintenance. I applied a recurrent neural network model to automatically classify stages of sleep of human participants based on physiological data read from EEG and EMG electrode recordings.

I've been tasked with studying memcomputing machines, Turing B-Type machines, reservoir computation, Hebbian learning, atomic switch networks, and how they relate to each other. The goal of my participation in the lab is to develop an overarching thesis on computation which is continuous in time and / or space, and involves some source of randomness.

Responsible for running participants on a sleep deprivation study, collecting and analyzing sleep and response inhibition task data, and presenting findings in a research paper at the semester's end.