Ariel Quinn Email & Phone Number

Mountain View, California, Us

Achievements:˖ Improved per frame recall from 20% → 80% for a critical underrepresented class by using SOTA model training techniques, without reducing precision˖ Accelerated labeling pipeline using a sign detector combined with OCR, reducing our data delivery time from months to days˖ Optimized, TRT’d, and deployed the model in C++˖ Managed labeling and metrics pipelines, provided feedback on cameras to the hardware team, coordinated labeling standards with other stakeholders˖ Led the detection of jersey barriers in an extremely timeline critical context. Delivered this P0 work on time, despite project being behind schedule prior to my joining.Technologies:Pytorch, torchscript, tensorboard

San Francisco, California, Us

Achievements:˖ Built a training and metrics pipeline for 3D semantic segmentation of lidar points; experimented with different architectures to improve precision and recall of the model˖ Deployed a lidar based semantic segmentation torchscript using C++ and established versioning standards to ensure model and metric reproducibility.˖ Implemented pipeline to deploy object detection results to a geospatial database and visualize in NodeJSTechnologies:Pytorch, torchscript, tensorboard

San Francisco, California, Us

Achievements:˖ Developed a low-latency feature matching algorithm in C++. This algorithm was a critical component of a localization tool my team developed to solve the kidnapped robot problem.˖ Worked with team to develop a structure from motion (SFM) pipeline to generate high definition 3D maps (HD Maps) at scale˖ Designed and led the implementation of a microservice which serves image based localization at scale with low latency and sub-meter precisionTechnologies:SLAM, SFM, and high performance algorithmsTools:COLMAP, OpenSFM, OpenCVKafka, HDFS, Docker, CadencePython, C++, Java

San Jose, California, Us

Achievements:˖ Extended the Element OS's API to a variety of technologies by creating code generators and creating the packages. These technolgoies included SDKs in Python, C#, Java, and Ruby. We also wrote a CLI using the same generator framework.˖ I set up continuous integration in Jenkins for said integrations so that every time a command changed in the Element OS, we could automatically push that change to all the downstream products. This entailed a great deal of cross team coordination.Tools:Python (primarily), C#, Java, and Ruby.Git, docker, jenkins

San Jose, California, Us

˖ Developed a flash cache simulator to test and train various machine learning prefetching algorithms at high speed which would be untenable with real data on a full storage array˖ Filed for a patent on a novel use of a machine learning algorithm for prefetching in a flash cache for a storage array given certain applications˖ Developed a ML algorithm for dynamically tuning storage array parameters based on workload characteristics˖ Worked primarily in R, Python, and C++˖ Applied Natural Language Processing techniques in a novel way and developed multiple prototypes of algorithms based on research papers

˖ Helped freshman design their first real engineering project. This entails attending class, helping keep them on track, holding office hours, helping them through unproductive slumps, and encouraging their creativity.

San Jose, California, Us

˖ Created a PowerShell library to manage block storage˖ Created an automated system in python for generating future updates to the PowerShell library by building on a Webservice REST API˖ Created an automated system for generating help docs, unit tests and other important artifacts˖ Created a windows installer to finalize the product˖ Used Python, C#, and Java during my internship

At the beginning of this internship, I learned to work with pneumatic actuators that my lab had already created. Then we decided to create a feedback system using a flex sensor and a pressure sensor. My job was to create an entirely new pneumatic actuator which would fit the flex sensor from sparkfun. Once I did that, I created a feedback system using the pressure sensor and an arduino. It allows a user to input a particular value for the pressure in psi and then holds the pneumatic actuator at that pressure value until a new value is entered. I am currently creating a PCB that will do the same thing, but using a different microcontroller.

Redmond, Washington, Us

The explorer interns rotate through the three primary positions in the company: the PM position, the Developer position, and the Tester position. Each explorer team is given one or more projects which they will see to the end. Our project was to improve the Code Coverage web tool for the Exchange team. During the PM phase, I met with various people to determine what the users actually wanted out of a tool like this and wrote specs. During the developer phase, we implemented and tested features for the tool. During the test phase, we worked on adding fields to a database collection and presentation tool.

I learned about computer architecture, the BLAS, compilers, using Janus (the supercomputer at CU Boulder), and Perl. While I was there, I wrote and ran some tests on the compiler my team had been working on.

Master Of Science - Ms, Electrical And Computer Engineering

Bachelor Of Science - Bs, Electrical And Computer Engineering

Frontiers Of Science Institute, Grounded Cognition