Jesse Miller Email and Phone Number

• Spearheaded development and deployed a PPO-based RL Planner (PyTorch, Stable-Baselines3, C++) surpassing safety and comfort benchmarks set by prior imitation learning based ML planner in sim and on-vehicle ACC scenarios.• Led a cross-org Reflexive Planner project, a low latency task for collision prevention critical to achieving the Driverless milestone. Coded features that reduced the false-positive rate by 400% (without impacting safety) anddesigned regression tests and metrics for performance monitoring. Reported results directly to Motional’s CTO.• Awarded Motional’s monthly “Mobilizer Award” and consistently received “exceeds expectation” for every performance review.

• Code-owner and developer for a nonlinear trajectory optimization based MPC formulated with FORCES Pro in Python and deployed in C++. Implemented dozens of features and gave thorough software and design reviews.◦ Collaborated to implement and patent a novel Contingency MPC approach for handling multi-modal predictions in trajectory optimizer, refactoring solver formulation to support joint optimization with minimal latency overhead.◦ Invented a novel algorithm for generating assertive lane change trajectories via “gap matching” behavior to signal intent. Redesigned the Planning/Control interface (C++) and added visualizations to support the new feature.◦ Tuned MPC weights and enhanced cost function, contributing to investor analysts’ feedback that the ride was“smoother than Waymo’s”. First tuned in open-loop sim to match human driving, later fine-tuning on vehicle.• Led Path Tracker Perfomance team (6 ppl) creating Python/SQL metric frameworks and designing automated release tests. Reduced rate of LQR’s high cross-track error events from 25 to 1 violation per 1K miles.• Collaborated to formulate and implement a convex trajectory optimizer (OSQP, C++, Eigen), linearizing objective and constraint terms. Reduced latency by 50% over NLP version.• Led Trajectory Formulation Team for a quarter; defined the team’s north star strategy, conducted one-on-ones, and oversaw successful demo deployment.• Developed and patented an ML based vehicle dynamics model for simulation (LSTM in PyTorch) and deployed it in C++, reducing average displacement error by 40% over 8s compared to existing physics based model.• Primary code owner and top contributor for repo containing Python analysis scripts used for AV log visualization and debug, regularly used by 30+ people. Adding log streaming support and improving deserialization speed by 100x.

• Owner of a dynamic bicycle model based path tracker module, refactoring and integrating code during the company merger. Migrating formulation from Acados to Forces PRO and refactoring code to follow SOLID principles.• Designed and coded “Mock Planner” and “Fault Injection” tools (C++, Python, IPC) used to stress test controller and for vehicle operator training.• Mentored intern who explored trajectory optimizer solver modifications for improved latency and performance.

• Worked on vehicle hardware integration and sensor calibration, outfitting the Renault Zoe EVs with nuTonomy’s AV compute stack. Troubleshooted and fixed issues such as an incorrectly wired GPS Serial cable using an oscilloscope.• Streamlined AV closed course testing, curating a weekly KPI report on AV performance (Python) shared with C-suite.

• Tuned a PID controller for a DC motor system using system identification and control design methods inMATLAB/Simulink.• Developed a electromechanical test evaluation setup using fiber optics, PXI, LabView, 3D printing, and control design to quantify a mechanism’s performance.• Designed tests and analyzed results to characterize the performance of various mechanisms relating to nuclear weapon safety.• Modeled and evaluated rigid multi-body equations of motion in 3D using Siemens LMS software.

• Created a simulation framework to model the 3D dynamics of a sailboat in MATLAB to determine the feasibility of autonomously sailing around the world.• Wrote a thesis paper describing a novel ‘rudder connected to sail’ design to improve the directional stability of a robotic sailboat. Presented findings at a robotics seminar with advisor Andy Ruina at Cornell University.• Provided recommended robotic sailboat design to CUSail project team at Cornell which they built and tested on Cayuga Lake, verifying the simulation results.

• Taught two recitation sections with about 15 students in each, held regular office hours, and graded exams• Volunteered to fill in for the Professor, teaching a one-and-a-half-hour lecture to about 150 students• Received 4.88/5 overall rating on my TA evaluation based on 16 student responses and was nominated to be a TA trainer

• Created a random forest ML predictor (Python, MySQL) to predict the time it takes to pick an item off a shelf.• Implemented the predictor service that improved metrics used to assess employee performance, saving the company millions and remaining in use years after my internship.

• Developed and wrote Python handlers to control a robotic ball, Sphero, with a path planning program, LTLMoP • Successfully demoed the code, having Sphero autonomously traverse a map and react to its environment

• Collaborated in a group to design and program in C and Arduino robot to autonomously battle other robots • Won several rounds and received an A+ in the course • Served as teaching assistant, holding office hours, grading tests, and overseeing the labs

• Optimized the rocket’s bulkhead design using SolidWorks and CES EduPack • Created in SolidWorks and constructed a tube cutting jig to cut carbon fiber tubing • Manufactured rocket parts in the machine shop using the mill, lathe, and laser cutter

• Invented a unique idea to listen to the sounds produced by handwriting and transcribe what was written.◦ Developed and coded a MATLAB program to transcribe handwriting by only using a microphone.◦ Won $15,000 in research funding by presenting my invention in a U.S. Air Force competition against 37 others.• Designed in NX Unigraphics and performed an experiment to test the effects of unsteady flow through an intercooler.• Published and presented a conference paper based on these findings in the SAE 2014 Aerospace Conference.