Joshua Manela

Working on a product-focuses team writing software for robots out in the field. If you have experience ... ... implementing learned behavior policies ... simulating robots ... developing ML infrastructure for robotic and / or embedding systems ... writing and deploying robotics middleware ... deploying perception models for safety-critical systems ... writing rust for robots please reach out! For more information, see our website https://bedrockrobotics.com/

Tech lead for various sensor fusion efforts across waymo. Projects ranged from IC contributions to management of projects across teams of roughly twenty-five to forty people. Managed and deployed Waymo's largest sensor fusion model, consolidating over a dozen nets into a single model. I still spent time training models and writing code while figuring out a hectic meeting schedule.

Responsible for the development, integration, and maintenance of novel long range sensor fusion algorithms. This lead to me spearheading Waymo's long-range sensor fusion strategy. I lead meetings, delegated projects, and presented to leadership while shepherding the work of over a dozen incredible engineers. I still somehow managed to write a ship a fair amount of code myself. Cross-functionally developed perception algorithms for next-generation radar products alongside hardware engineers. Collaborated with the research team to push the state-of-the-art in long range sensor fusion.

As one of the original members of the team, I worked across the company designing novel perception algorithms, building low-latency C++17 infrastructure, and pushing several features to the car. I worked on projects spanning mapping, lidar processing, image detection, geometric computer vision, tracking, and core runtime infrastructure. I spent a fair amount of time at argo leading the development and integration of a deep lidar object detector. This included getting labels, working with labeling companies, cleaning data, writing modeling code, optimizing the model with TensorRT, deploying the model onboard, integrating the output with other teams, and metrics throughout each step of the process (and after!) This then grew to my first tech-lead role of managing the day-to-day projects of around eight people. To further improve Argo's lidar capabilities, I worked with the Argo Lidar team enhancing the firmware capabilities of our custom geiger-mode lidar sensor. I also performed novel published research in stereo vision.

Trained models and wrote C++ infrastructure for solving end to end self driving problems. Wrote onboard software for long-horizon actor prediction.

Started as an intern working on the "Machine Learning Team" collaborating with various other groups throughout the company helping to incorporate machine learning into classical robotics problems. Beyond helping other teams, I owned a large portion of an end-to-end (image to steering wheel & pedals) deep learning model. Although originally this was meant as an internship between my undergraduate and PhD studies, I ended up not pursing a PhD and transitioned to being a full-time employee.

I tutored in both the Electrical and Computer Engineering (ECE) and the Computer Science (CS) Learning Centers as a student coach to mentor students through related courses. ECE Courses Tutored: Circuits I & II, Electronics, Linear Systems, Digital Logic, and Math for ECEs CS Courses Tutored: Intro to Programming I & II, Data Structures, Systems Programming, Discrete Structures and Artificial Intelligence.

Worked under Dr. Timothy C. Havens in the Pattern Recognition and Intelligent Machines Engineering Lab. Investigated various Simultaneous Localization and Mapping (SLAM) algorithms for use on autonomous drones. Designed and fabricated a data collection system with a planar lidar, a gig-e blackfly camera, and a six axis IMU that was mounted on a drone. Implemented camera-lidar SLAM sensor fusion algorithms. Designed, wrote, and published a novel extension to Particle Swarm Optimization (PSO).

Mentored students during two lab sections of the Intro to Programming Class (CS 1121) at Michigan Tech.

Responsible for the well being of roughly forty undergraduate students. Planned and organized hall activities and programs Made informative bulletin boards every month Monitored the hall and responded to situations in the dorm to get them under control Wrote weekly fire reports and monthly block reports Wrote concern reports, warning reports, and incident reports as necessary

Developed hardware power tests for switches deployed across several data centers. Influenced the design and manufacturing decisions of custom network switches.

Implemented RESTful APIs which posted microservice metrics to a custom NoSQL database to monitor for suspicious activity.

Designed, Tested and Built an encoder simulator box which allowed the user to select a desired RPM, CAM phase, and Encoder Type. The box outputted a half-moon CAM signal, VR+- (analog encoder), as well as a Digital Encoder Signal. Wrote drivers for the Mercury Marine Security System, specifically the CAN, RFID, and general IO libraries. Interpreted the encoding coming out of the RFID key fob and generalized everything so improvements can be made in the future. Wrote code to convert specific SmartCraft Fault messages to J1939-DM1 messages. Wrote scripts to automate testing of multiple faults.

Set up industrial robotic cell layouts for various integrators and manufactures. Used the Delmia IGRIP CAD systems to simulate robotic assembly lines. Programmed macros for Delmia IGRIP to make use of the software more efficient.

Taught students the fun of building with Legos and designing/ programming Lego robots. Designed curriculum for various robotics camps, classes, and events. Wrote the framework code for the current "SUMO" system in use. Helped students connect large lego cities by simulating a real-life power grid.