Hi, I'm a computer science student at the University of Washington. I love writing code for physical platforms, things I can touch and can affect the real world. I consider myself a creative who loves to bring things off the page and into real life.

ResumeContact me
scroll

    • The system lacked a way to process a key category of incoming sensor data in real-time, creating a gap its analysis pipeline.
    • Designed and built a multi-threaded microservice in Rust from scratch to perform thousands of time-sensitive calculations per second on incoming data.
    • Architected a gRPC API to integrate the service with the broader analysis system, and implemented the algorithm that combined its output with live data to generate real-time assessments.
    • Validated the system using an in-house simulator generating synthetic data, confirming reliable performance under high volumes of data.

    • Migrating the LVBMS PCB from the LTC6813 to the ADBMS6830 required a complete firmware rewrite to support the new analog cell monitoring IC
    • Designed low-level SPI drivers in C++ to configure voltage, temperature, current monitoring, and passive balancing on the ADBMS, and to transmit telemetry over CAN via the MCP2515 transceiver.
    • Implemented RTOS tasks to poll the ADBMS and MCP2515 concurrently, and to detect faults in real-time
    • Performed hardware in the loop testing with a seven-cell Li-ion pack, cross-checking voltage and temperature readings against a voltmeter and ground-truth sensors, verifying CAN communication, and injecting faults to confirm the protection logic fired.

    • All health insurance policy documents are public, but unstructured, and the company needed to aggregate it all in a structured data format.
    • Designed an LLM powered pipline to extract structured data from insurance company websites, PDFs, and other unstructured text sources.
    • Benchmarked multiple LLMs and prompt strategies to find the most accurate and cost-effective solution for extracting structured data from unstructured text.
    • Integrated LangChain and Pydantic to strictly enforce JSON schema validation, eliminating malformed outputs by 100%.

    • Existing rover localization and pathfinding algorithms rely on satellite-generated digital elevation maps but many extraplanetary environments lack satellite coverage.
    • Benchmarked monocular depth-estimation models from Hugging Face with the KITTI dataset on a Jetson Orin Nano GPU, and selected the DepthAnything ViT for the rover localization algorithm.
    • Conducted a literature survey of existing cost based pathfinding algorithms and developed obstacle/elevation scoring algorithms to determine traversable areas.
    • Accelerated image-to-map conversion by running PyTorch depth inference on a Jetson Orin Nano GPU with CUDA.

    • Relevant coursework: Discrete Mathematics; Linear Algebra; Computer Architecture and Hardware-Software Systems; Probability, Statistics & Machine Learning Foundations.

Autonomous Car

2026

Augmented an RC car chassis with my Jetson Nano running ROS2 to navigate the sidewalk.

PythonROS2Jetson Nano
In progress

Overwatch Controller

2025

Point and shoot like you're actually in the game

C++Arduino3D Printing
GitHub

Cost based Path-Planner

2025

A monocular camera based approach to rover localization and path planning.

PyTorchMLJetson Orin Nano
GitHub

Internia

2024

A web app that helps users find research papers with a like/dislike reccommendation system.

TypeScriptReactMongoDB
GitHub
aravjaiman@outlook.com

Drawn and built by hand — Arav Jaiman