Teaching

HarvardX: Tiny Machine Learning

Teaching Staff LeadFall 2020 and ongoing
  • Co-designed a new MOOC specialization in three 5-week courses on the EdX platform covering the emerging field of Tiny Machine Learning (deploying machine learning onto microcontrollers for machine learning at the extreme edge) with the aim of democratizing access to this developing field
  • Over 5,000 students enrolled since the September launch
  • Managed the 9-person course staff to ensure that content was created, reviewed, and produced in a timely manner
  • Led and managed external relations for the course team coordinating with edX, Google, and Arduino
  • Co-designed course materials including short video lectures, readings, code walkthroughs, assessments, and discussion forums

Harvard CS 249r: Special Topics in Edge Computing: Tiny Machine Learning

Head Teaching Fellow (Head TA)Fall 2020
  • Co-designed a new 40 student course on the emerging field of Tiny Machine Learning (deploying machine learning onto microcontrollers for machine learning at the extreme edge)
  • Designed and gave lectures for the introduction to machine learning section of the course
  • Co-developed hands-on project-based assignments training TinyML models with Google Colab and deploying on Arduinos
  • Mentored student teams pursuing research-based final projects

Harvard CS 249r: Special Topics in Edge Computing: Autonomous Machines

Head Teaching Fellow (Head TA)Fall 2019
  • Received the Derek Bok Center Distinction in Teaching Award
  • Co-designed a new 45 student course at the intersection of robotics and computer architecture / systems serving as the robotics expert and instructor
  • Designed and gave lectures for the robotics section of the course
  • Co-developed course assignments and course infrastructure/tools (e.g., the online paper discussion forum)
  • Mentored student teams pursuing research-based final projects

Harvard CS 182: Introduction to Artificial Intelligence

Head Teaching Fellow (Head TA)Fall 2017, 2018
  • Received the Derek Bok Center Distinction in Teaching Award in both 2017 and 2018
  • Ran a team of 11 teaching fellows to ensure that: sections and office hours were held, exams and homework assignments were graded, and student questions on the online forum were answered in a timely manner
  • Designed and gave two lectures: ‘Introduction to Robotics and Path Planning I/II’
  • Aided in the development of course assignments, and course infrastructure/tools (e.g., autograders)
  • Mentored student teams pursuing research-based final projects

MIT MAS.863: How to Make Almost Anything

Teaching Assistant for the Harvard SectionFall 2017, 2018, 2019
  • Led introductory sessions for various pieces of software and hardware used in the course (e.g., embedded programming for Atmel microcontrollers, CAD in Solidworks and Eagle)
  • Held office hours, aided students in lab work, machine usage, and project design

MIT Beaverworks Summer Institute: Autonomous RACECAR Grad Prix

Associate InstructorSummer 2016, 2017, 2018, 2019
  • Worked with 9-12 teams of 4-6 students to teach programming concepts and robotic algorithm design through the completion of autonomous tasks in a Python/ROS environment using the MIT RACECAR hardware platform
  • Co-designed weekly challenges to ensure all teams developed the technical skills needed for the final race
  • Co-designed and co-built the final race track spanning an entire ice hockey rink

Harvard Math 1a/b: Introduction to Calculus I/II

Course AssistantFall/Spring 2010
  • Taught a weekly breakout section, staffed the Math Question Center to aid students in problem sets, and graded problem sets

Featured Publications

In this paper, we detail the designs of three faster than state-of-the-art implementations of the gradient of rigid body dynamics on a CPU, GPU, and FPGA. Our optimized FPGA and GPU implementations provide as much as a 3.0x end-to-end speedup over our optimized CPU implementation by refactoring the algorithm to exploit its computational features, e.g., parallelism at different granularities.
PDF Code Publication

In this extended abstract we extend our previous work by using our Parallel DDP implementation for MPC on a physical Kuka arm. We demonstrated the feasibility of this approach in the presence of model discrepancies and communication delays between the robot and GPU and found that higher control rates generally lead to better tracking performance across a range of parallelization options.
PDF Code Poster Video

We analyze the benefits and tradeoffs of higher degrees of parallelization using a multiple-shooting variant of DDP implemented on a GPU. We describe our implementation strategy and present results demonstrating its performance compared to an equivalent multi-threaded CPU implementation using several benchmark control tasks. Our results suggest that GPU-based solvers can offer increased per-iteration computation time and faster convergence in some cases, but in general tradeoffs exist between convergence behavior and degree of algorithm-level parallelism.
PDF Code Slides

We build on recent work on Unscented Dynamic Programming (UDP) - which eliminates dynamics derivative computations in DDP - to support general nonlinear state and input constraints using an augmented Lagrangian. The resulting algorithm has the same computational cost as first-order penalty-based DDP variants, but can achieve constraint satisfaction to high precision without the numerical ill-conditioning associated with penalty methods.
PDF Code Poster Slides Publication