Independently developed AI-Powered CoPilot for Cisco Catalyst Center Template Editor. Built interactive, stateful, multi-turn RAG system for dynamic CLI auto-generation, streamlining template development.Led design and development of various features for Cisco's IoTOD Secure Equipment Access service including multi protocol device access, third party account integrations, DUO multi factor authentication, WebEx screen monitoring and recording, session history auditing, mobile device support, event-based notifications, bulk system configuration, high availability / scalability support, session metrics, request based access workflow, cross-tenant remote access brokerage, and self-updating binary application.Performed and presented robust analysis of industrial remote access to identify target markets, potential growth areas, competitive advantages, and valuable feature additions.

Over the course of three summers, I actively contributed to various aspects of Cisco's IoT Edge Intelligence Project, working alongside full-time employees, following the Agile Scrum development process, and directly integrating my work with Cisco's pre-existing platforms.2021:Constructed an interactive Docker swarm framework for simulating Trinity Box topologies, helping enable Cisco to test device capabilities at scale before product release. Creatively automated deployment using SpringBoot cloud functions and IOTD API calls. Secured all inter-service communication with Docker secrets and Rabbitmq messaging. Scaled performance immensely, capable of launching thousands of virtualized routers.2020:Added fully-interactive, cloud-powered debugging to Cisco's VS Code Edge Intelligence Extension, allowing developers to remotely attach to data logics running on edge gateway devices (i.e. Cisco routers) and debug the scripts in real-time from their local workstation. Capabilities include setting breakpoints, subscribing to variables, visualizing the stack frame, and streaming standard output and error. Secured the debugging connection using an AWS rendezvous server, multiple SSH tunnels, and a custom SpringBoot credential-generating microservice. Automated connection details to ensure a seamless user-facing experience.2019:Created an operations tool for machine developers to easily transform complex data streams from connected devices into well-structured schema. Capabilities include constructing a custom output schema, mapping input to output through a simple drag-and-drop interface, and applying various data transformations. Integrated the tool as a precursor for Cisco's Edge Data Services Platform (EDS). Developed connectors for various machine protocols including MQTT, OPC UA, and Modbus. Achieved 80% code coverage with fully automated unit testing.

Co-lead DevX's BMaps (formerly Mappening) project, a web-based application that aggregates UCLA events from a variety of sources onto a single, interactive map/calendar interface. I led meetings, delegated tasks, reviewed code submissions, spearheaded recruiting efforts, fixed bugs, added new features, drove design changes, optimized site performance, and ensured cross-platform compatibility while working as a full-stack developer and project lead.

Independently theorized and developed a Java algorithm to autonomously extract demographic information from PubMed articles and assign MeSH search tags. Implemented, trained, and optimized a host of ML classifiers using Weka. Achieved 99% recall and 75% accuracy using a combination of five classifiers. Integrated NLP methods to programmatically parse relevant information from XML-formatted tables. Code currently deployed in production as part of the Selective Indexing Initiative within the National Library of Medicine.

Contributed to the Project Sidewalk research project, operating under the Human-Computer Interaction Lab at University of Maryland, College Park. Virtually mapped the street level accessibility of Washington D.C. to ease navigation for those with ambulatory disabilities, achieving 100% auditing coverage of Washington D.C. sidewalks (1,075 miles). Implemented front end features using JavaScript, HTML/CSS, PHP, jQuery, and Bootstrap. Visualized user activity by combining PostgreSQL queries with D3 JavaScript graphs. Independently designed and developed a versatile ground truth resolution tool using JavaScript in conjunction with various GIS-related APIs such as Mapbox, Google Maps, and Google Street View. Performed in-person user field research as well as in-depth statistical analysis of MechanicalTurk user interaction metrics in order to develop early-stage low-level quality control mechanisms. Co-authored a research paper "Project Sidewalk: A Web-Based Crowdsourcing Tool for Collecting Sidewalk Accessibility Data At Scale" for the 2019 ACM CHI Conference.