Resume

Honors & Awards:

• HDTV Grand Alliance Fellowship [2026] — selected for funding award, given exemplary academic and research achievements

Research:

• Medical AI @ Katabi Lab [2026–]

Research:

• Machine Learning @ Jha Lab [2025–2026] — first author, lead ML contributor, and primary writer of papers on diabetes detection
  • Publication: SweetDeep: A Wearable AI Solution for Real-Time Non-Invasive Diabetes Screening
  • Publication: Machine Learning for Type 2 Diabetes Prediction using Continuous Glucose Monitors
• Robotics @ Silver Lab [2025–2026] — improving real-to-sim-to-real transfer through perception and simulation research
  • Internal Research: Optimizing Robust Object-Centric Prompts for VLM-Based Robot Perception
  • Internal Research: Building a Kinematic Simulator for the Princeton Robot Planning and Learning Lab

Outreach:

• Robotics @ Silver Lab [2026] — built a Scratch-like website, then utilized it to teach robotic abstractions to middle-school students

Honors & Awards:

• Class of 1939 Award [2024] — honored for the single highest academic standing among 1300+ members of Princeton's Class of '25
• James Hayes-Edgar Palmer Prize [2025] — recognized for excellent scholarship, leadership, and creativity in engineering work
• G David Forney Jr Prize [2025] — awarded for an outstanding record in communication science, systems, and signals coursework
• 2x Shapiro Award for Outstanding Academic Achievement [2022, 2023]; Manfred Pyka Memorial Physics Prize [2022]
• 1st Place in Harvard's Pacbot Robotics Competition [2023]; 2nd Place in the Princeton Computer Science Contest [2023]

Leadership:

• President, Mentor, Software + Firmware Lead on Princeton Robotics Team [2023–2025] — led Pacbot and Golf Cart teams
• Campus Service Leader of Loaves & Fishes [2023–2025] — volunteering to feed homeless and low-income residents

Research:

• Robotics @ Fisac Lab [2024–2025] — developed an ILQR-based self-driving approach to refine obstacle avoidance on model trucks
• Machine Learning @ Jha Lab [2025] — trained accurate, compact, and well-calibrated smartwatch disease detection models
  • Undergraduate Thesis: Real-Time Disease Detection with Energy-Based Models
  • Internal Research: PreCEPT: Pretraining Classifiers using Energy-based Priors and Teachers
• Hardware @ Martonosi Lab [2024] — designed and verified a pipelined accelerator for chaining frequent priority queue operations

• Burnett Honors Scholar

• Class Rank 1 of 1030
• Valedictorian
• Physics Olympiad Qualifier

• Completed and deployed an end-to-end service for translating and formatting Google Slides decks using Vision-Language Models, with built-in version control, carefully designed billing, and lazy formatting updates upon changing content
• Integrated GitHub and Slack webhooks for a code localization agent, accelerating the developer feedback loop
• Hypothesized and validated edge-case vulnerabilities within the agent's billing pipeline, correcting them to prevent under-billing
• Architected an automatic validation pipeline to benchmark and patch regressions in natural language translation quality
• Expanded the suite of supported file formats for translation—including Lottie and SVG—using vision-language agents

• Collaborated with clinical team to process and synchronize wearable sensor data streams for disease classification dataset creation
• Engineered and validated a new approach to train transformer and multilayer perceptron models on physiological timeseries data, while also refining calibration and improving consistency of model performance across training runs
• Constructed and selected clinically relevant features, then trained 100x smaller classifiers with 10+% higher accuracy and F1 scores
• Tripled model iteration speeds by automating parallel multi-fold training and evaluation within existing CI/CD infrastructure
• Coordinated the integration of machine learning screening models into smartwatches, using C++, Java, and Android Studio

• Built a Python library for training energy-based model classifiers across diverse (tabular & physiological) disease detection tasks
• Achieved state-of-the-art accuracy on multiple proprietary clinical datasets using a novel energy-based model distillation approach
• Enabled out-of-distribution detection and semi-supervised learning in existing tabular datasets via energy-based modeling
• Conceived and mentored an undergraduate research project using calibrated models for differential diagnosis applications
• Supported data processing and model training for a post-doctoral project that predicts emotions using physiological signals

• Led tutoring, assignment help, lab design, and teaching for nine courses: Physics, Advanced Vector Calculus, Advanced Mechanics, Programming Systems, Algorithms, Logic Design, Circuits, Autonomous Systems Lab, Intelligent Robotic Systems
• Assisted over 200 students with problem set questions, labs, debugging code, and exam preparation

• Introduced a framework for automatic profiling, performance analysis, power analysis, and visualizations of runs across multiple data center cluster workloads, and fully integrated it into the team's standard set of tools
• Devised and deployed an Ansible tool which automatically tracks and visualizes usage of shared testing boards
• Planned and launched a new CI/CD pipeline to catch bugs introduced into a widely used internal benchmark testing suite
• Tested and enhanced a kernel driver that uses Linux tracing to sample profiling events for fast workload characterization

• Augmented an interactive and comprehensive cloud-based kernel event analysis tool, used by automotive engineers and customers to identify ways to improve startup and runtime efficiency of OS and virtualization layers on the Drive Orin
• Automated the process of diagnosing common performance issues by constructing a pipeline to analyze OS trace logs
• Debugged and improved a real-time lightweight utility for estimating CPU and memory utilization in self-driving cars

• Conducted exploratory research of power measurement approaches for the Grace CPU superchip
• Studied, modified, and wrote kernel device drivers, Device Tree bindings, ACPI table entries, and bash scripts to configure and process power sensor inputs to the chip to enable power telemetry and monitoring
• Added enhancements to a power analysis tool, allowing for interactive visualizations of large traces of power and thermal data

• Implemented testing libraries in C and C++ to evaluate and summarize the real-time performance, thermal, and power metrics of embedded NVIDIA Quadro GPUs, PCIe connections, and the company's graphics cards
• Demonstrated graphics and multiprocessing capabilities of hardware through TensorFlow and OpenCV applications