A timeline of how I arrived at my current state of knowledge, from where I am now to my very beginnings in software engineering.
I'm currently building Nvidia's learned driving policy for their L2++ product. I'm excited to be back in the thick of difficult technical challenges, and proud to contribute to one of the most influential companies in tech.
After investing time in personal knowledge I was ready to apply it to industry problems like generative AI and robotics.
In early 2023 I joined Spanning Labs, a SaaS company powered by AI that helps companies with documentation. I researched and built the first version of their core product, known internally at the Spanning Wizard. I maintain an advisory role.
This past year has been an integration of all the knowledge gained from my 10 years of working in industry.
My intention was to build my skills in novel machine learning techniques, and I accomplished this by publishing several open-source projects.
I also wanted to step back and consider my goals for the next decade of my career and life.
In 2022, I traveled to Nepal, India, Thailand, Indonesia, and Kyrgyzstan, where I got to trek across the Himalayas, and scuba dive with manta rays.
Around the beginning of 2020, Zoox was acquired by Amazon.
An opportunity for long-term, future-forward projects emerged. I pivoted into machine learning, AI, and reinforcement learning.
I investigated learning-based approaches for decision making, such as how to follow specific rules of the road. This included stopping for stop signs, following others, and yielding at intersections.
In 2019, I made the enthusiastic decision to step back from my lead role in collision avoidance to join Zoox's Research Team.
I wanted to see what the forefront of the planning and control space looked like, and that meant doing research full time.
I worked on integrating new forms of machine-learned prediction into dense tree searches for planning.
I created an on-vehicle demo for executives. It was highly successful – the new method was way more adaptable and behaved more naturally. My findings influenced how the entire Planner team developed features.
After a year of working on highway behavior algorithms, I led a new collision avoidance system (CAS) team. This output of this project is a backup system that double checks the output of the AI and ensures it's collision-free. I took the project from a one-file Python prototype to a validated ROS component running on the entire fleet.
To demonstrate success, I had to increase the maximum deceleration of the fleet. This involved hacking into Toyota CAN signals and brake potentiometers so we could fake signals to manipulate the vehicle.
As part of the validation, I conducted testing with fake pedestrians and incredible views of San Francisco.
In 2017, I moved from Zoox's Firmware team to the Planner team so I could develop motion planning algorithms.
My work started with learning how to stop for jaywalkers, and progressed to stopping for crosswalks by anticipating pedestrian light cycles. I also worked on yielding at intersections by doing geometric collision checks. The yielding logic evolved and grew to include vehicles going out of turn at stop signs and yielding to bicycles when traversing over bike lanes to make turns.
Over my entire time at Zoox I was issued over 10 patents.
Later in the year, I led the new high-speed highway driving effort. I managed all issues and ensured they got handled by working with other engineers, QA, and vehicle operators. I personally worked on zipper merges and using costs to determining which lane to be in.
The project resulted in a 100% fully autonomous demo to Bloomberg where I was the safety driver.
After graduating my masters program in Mechanical & Aerospace Engineering, I joined the self-driving car startup Zoox.
I began as a Firmware Engineer and wrote the safety-critical code that ensured the AI stayed powered on, and that the safety driver would remain in control of the vehicle.
This involved writing requirements via systems engineering, programming HIL testers, writing software, and validating that all components and features worked together successfully.
In 2015, halfway through graduate school, I took a hiatus to rejoin Tesla for a second internship. My first one was in supply chain logistics, and this one was on the Autopilot team. I quickly became part of the core group pushing Autopilot v1 out the door. I traveled frequently to LA to test new changes on Elon's 405 commute into SpaceX.
I personally worked on integrating the in-house learned fleet maps into the lane-keeping algorithm to stay centered on the lane. This was super important when other modalities failed, such as when the car's cameras were no longer seeing lane lines that might be unclear.
I decided to return to UC Davis for my master's degree in engineering. Working in manufacturing at Rolls-Royce Motor Cars reaffirmed my deep passion for software, controls, and motion. I worked harder than ever to crystalize my career path in this direction.
I developed projects such as programming a tractor to autonomously avoid obstacles, implementing a Kalman filter to estimate relative position, classifying restaurants with a convolutional neural network, and designing a satellite to reboost the Hubble Space Telescope in orbit.
I worked as a Manufacturing Engineer at Rolls-Royce Motor Cars. I loved the diverse experience of living in a different country and understand how well the world works with different norms.
I learned how to optimize the manufacturing experience for associates in low-volume settings. I also learned to appreciate and create clear visual communication for presentations and documentation.
I landed an internship at Tesla! The purchasing team wasn't my first choice, but I was honestly excited to be part of the team and see the first Model S vehicles come off the line.
I did a variety of oddball tasks including creating a internal supplier score card to evaluate supplier performance, and ultimately decide when to upgrade them or phase them out. I also created work instruction videos for how to manufacture parts for the Model S.
My interest in dynamics and control was fostered during my undergraduate education. I truly enjoyed observing how robots move through the world and implementing control theory to influence how they move.
To practice the theory I learned in my courses, I joined Baja SAE, where I built a single rider off-roading dune buggy to compete against others in a 4-hour endurance race.
The year I joined was the first year the school ever competed. We over-engineered the buggy, and weren't able to pass technical inspection or compete. It was a sad year.
The following year I became captain, and we were fully in the zone to compete! We simplified the entire vehicle. I designed and fabricated the drivetrain. With some hard work and elbow grease, we passed technical inspection and were able to compete in the endurance race. Everyone was estatic!
My high school computer science class was where I first started programming. I was amazed by the power of software, and how you could control a computer to do so many things.
I made a tower defense game as part of my class, and was humbled to see how everything fits together.
