Uma Phatak Email & Phone Number

California, United States

San Francisco Bay Area

LLM Security

Stanford, California, United States

• Developed a new ethics curriculum as Embedded EthiCS CA for CS 124: From Languages to Information, with strong positive feedback from students• Developed new attendance app in PHP and Javascript for CS 278: Social Computing• General TA for CS 221: Principles and Techniques of AI• Hosted weekly 1:1 and group office hours, teach sections, and review and grade homework

San Francisco, California, United States

• Built a toxicity classification model to deploy in AI Firewall, reducing dependency on third-party models

• Designed and taught a 5-week, in-person Intro to Python curriculum from scratch, including creating homework assignments, assessments, and classwork. Taught in Google Colab.• Taught daily for 4 hours and hosted extra homework help sessions • Mentored underrepresented rising high school sophomores with limited CS background

Stanford, California, United States

• Taught weekly section of ~10 students through the Intro to CS course material (CS 106A, CS 106B, and ACE classes), grade coursework, and provide code debugging help.

Stanford, California, United States

• Web scraped in Python and performed statistical analysis on US IT-blog discourse surrounding Section 230, to identify key stakeholders in relation to the law• Compared US discourse to discourse around Germany's NetzDG internet speech law• Assembled and acquired metadata for a US IT-blog corpus and Section 230 subcorpus• Worked with Poetic Media Lab, in conjunction with the Center for Spatial and Textual Analysis (CESTA) at Stanford

Los Angeles, California, United States

• Worked on Google Lens' Visual Shopping AI Team (VSAI); introduced in-video object tracking, to improve understanding and accuracy of product detection in YouTube videos

Remote

• Worked on 3 projects with the engineering team for 4 months, to implement features to help the Media, Econometrics, and Client Services teams• Planned and implemented a Streaming Experiments page to ascertain the efficacy of previous TV campaigns; wrote backend Java job to execute an existing SQL query; wrote backend APIs in JSX/TSX to start the job and parse the returned data; designed frontend interface to take in intuitive inputs and populate the page with accordions for each experiment.• Helped create preliminary media plans for new clients, based on media plans for existing clients; wrote backend API in JSX/TSX to output a weighted average of existing companies' KPI (key performance indicator) data; started to modify existing linear optimizer in Python to use this synthetic data as input for a new client• Created page to display existing linear canonical networks; wrote backend GET and POST APIs and used an existing frontend template to display the data

Remote

• Worked with the nonprofit's engineering team to implement and test features on the Empower App, to improve the experience of volunteers and/or organizing directors who use the app for relational organizing• Frontend: used Javascript and Typescript to implement features like like logging who invites volunteers in history, and porting over code in React Native to React Native for Web• Backend: learned and used SQL to interact with the database

Remote

• Worked with NY-based nonprofit Terra Firma to make website more patient- and donor- facing, focusing specifically on enhancing branding and improving UX, to reach over 500 adjusting US immigrants per year• Learned Figma for website prototyping• Improved ease of engaging online with the 35 medical, legal, and educational services of Terra Firma

Laurel, Maryland

• Developed terrain navigation algorithms to estimate yaw measurements and model, detect, and mitigate dead pixels for NASA's Dragonfly mission (Dragonfly spacecraft will reach Saturn's moon Titan in 2035)• Analyzed mathematical and function-based methods of suppressing noise in light curves using Finite Impulse Response filtering for a Missile Defense Agency (MDA) project• Verified accuracy of an MDA project by running 7 simultaneous regression algorithms and a neural network on relevant database

Remote

• Learned React.js to help port HTML site to Gatsby.js• Updated website content including translations, cookie consent banners, and SEO errors, using HTML/CSS and JavaScript• In affiliation with Stanford COVID-19 Response Innovation Lab