Unnathi Utpal Kumar Email & Phone Number

Mountain View, Ca, Us

Atlanta, Georgia , Us

Language, Intelligence, & Thought (LIT) Lab (January 2024 - Present)• Awarded the President's Undergraduate Research Award (PURA) • Spearheading latent factor analysis and hypothesis-driven evaluations of LLM performance on the EWoK benchmark to enhance interpretability and explainability, focusing on word frequency, sentence length, and NLI concepts• EWoK Paper (arXiv): https://arxiv.org/abs/2405.09605v1Contextual Computing Group (January 2023 - Present)• Led the design and implementation of open-source Android and iOS libraries for ASL recognition with Google MediaPipe• Developed Python automation scripts to process anomalous video signing data (210000+ videos) & annotated ASL videos for Google’s Sign Language Recognition contest (NeurIPS 2023) in collaboration with DPAN and NTID (RIT)• Open-source Android Project (SLR-GTk-android-core): https://github.com/Accessible-Technology-in-Sign/SLR-GTk-android-core• Open-source iOS Project (SLR-GTk-iOS-core): https://github.com/Accessible-Technology-in-Sign/iOS-ToolKit• NeurIPS 2023 Paper: https://proceedings.neurips.cc/paper_files/paper/2023/file/00dada608b8db212ea7d9d92b24c68de-Supplemental-Datasets_and_Benchmarks.pdf

Atlanta, Ga, Us

UTA for CS1331: Intro to Object-Oriented Programming in Java during Spring 2023, Spring 2024 (Recitation Lead), Fall 2024 (Autograder Lead)• Conducted weekly office hours and recitation lessons to instruct and assist 30+ students in object-oriented programming in Java and debugging under Dr. Aibek Musaev and Richard Landry (Spring 2023, Fall 2024) and Dr. John Stasko (Spring 2024)• Created course materials like programming exercises, homework assignments, and recitation documents• Graded weekly programming assignments and midterms for 90+ students considering code quality and correctnessFrom student evaluations: • "Unnathi was extremely helpful in clearing my doubts during office hours and was really supportive of my learning journey through CS 1331. She is extremely knowledgeable of course concepts and communicates this knowledge effectively with students."• Greatest strengths: "communication", "knowledge of content"

Sunnyvale, Ca, Us

• Interned in LinkedIn's Model Cloud team (Model Serving Infrastructure) within Machine Learning Infrastructure• Automated the migration of LinkedIn's model serving stack from REST to gRPC and Kubernetes by developing a user-facing command-line tool (using Python and Click) deployed on all LinkedIn machines and production hosts, thereby reducing latency by 30% and recovering 1/2 FTE support/on-call• Enabled client-side re-routing (in a Java-based package) for inference using a gRPC-based stack and implemented support for parity checks between old/new stacks, requiring minimal effort from users for integration, parity testing, and ramping

Atlanta, Ga, Us

Project Lead for Fall 2023 - Spring 2024• Led a team of 15 students in collecting data for and developing an ML model that predicts job scams• Researched and applied ML techniques for fraud detection, culminating in an MVP that enhances detection accuracy and presents a user-friendly platform

Atlanta, Ga, Us

• Researched Intelligent Digital Communications under Dr. Ed Coyle to develop algorithms for cognitive radio• Developed a convolutional neural network to recognize and label drone and bluetooth signals in wireless signal spectrograms using PyTorch and TensorFlow, achieving 94% accuracy• Led quantization, pruning, and knowledge distillation efforts to optimize CNN for FPGA deployment and reduced resource use

Atlanta, Ga, Us

• Developed production-ready features and components (including a demoed platform-wide tree view with Storybook documentation and modern APIs) as part of a product migration to React from AngularJS and JSP (full-stack engineering team)• Created and upgraded 30+ unit tests using Cypress, Selenium, and Jest, improving code coverage upto 95%

Houston, Texas, Us

• Designed and built an open-source Kubernetes operator (using KubeBuilder) in Go to automate the configuration, deployment, and monitoring of a SPIRE server and agents in a cluster to implement SPIFFE standards in the GreenLake Cloud Platform• Developed custom resource definitions (CRDs) for components and comprehensive unit tests using GoMock• Implemented support for time-to-live (TTL) mechanism for SPIFFE Verifiable Identity Documents (SVIDs) in JWT format alongside other open-source contributions in the SPIRE Controller Manager• Open-source project: https://github.com/HewlettPackard/spire-k8s-operator

Champaign, Il, Us

• Delivered instructional lessons on computational thinking and the Wolfram language to middle school students• Supervised advanced projects in the Wolfram language, with topics ranging from machine learning, linguistics, and mathematics, to completion with individual attention to some students

Beverly Hills, Ca, Us

• Developed a cross-platform mobile app, with music streaming services and authentication, using React Native and Firebase with TypeScript • Collaborated on the app with 3 undergraduate interns from Cornell University• Assessed and troubleshot bugs on the web app, built using React and JavaScript

Cochin, Kerala, In

• Trained a machine learning model to predict vehicles in LiDAR point cloud images based on the VoxelNet architecture for use in autonomous robotic components• Analyzed and identified publicly available datasets with vehicle annotations through academic papers

Bachelor Of Science - Bs, Computer Science, Minor In Economics

High School Diploma

Wolfram Emerging Leaders Program

Wolfram High School Summer Camp