Christopher Correa Email & Phone Number

- Signals and Strategy Research — Developed processes to fit models on market movements. Explored the effect of sampling schemes, objective labels, model selection criteria, and meta strategy parameters on backtest simulations and live performance. - Data Pipeline — Designed and owned a system to parse market events for live trading strategies or capture market events for model training. - Infrastructure — Designed a nightly AWS ETL pipeline to format multiple exchanges' data to efficient formats for model training that could also be used to backfill historical data. Managed infra to remotely manage and deploy strategies.

San Francisco, California, Us

- Drove the technical roadmap for the team alongside a Product Manager- Managed a group of 7 engineers- Managed projects like item bounding box detection, item barcode scanning, oriented placement into a bagging machine, and robot and operator UIs.

San Francisco, California, Us

- Robot Policy Development — Re-architected the system design for all robot motions and policies to allow arbitrary policies to be chained together. Designed pick policies for suction cup grippers to intelligently approach items quickly without damage, increasing overall system cycle time by 240%. - Published a US patent (US20210081791) as a result of this work- Data Integrity — Developed various techniques to detect faulty sensors, obsolete camera calibration, etc. Created a data pipeline for per pick analytics and real time sensor streams. Created dashboards for per-SKU performance and analytics scripts to make informed policy decisions based on the recorded sensor data.- Testing — Developed an environment to simulate hardware and test robot policies for faster development without use of physical hardware and testing software on customer environments before deployment.

Berkeley, Ca, Us

I have taught EE 106B/206B (Robotic Manipulation and Interaction) for two semesters. The material in EE 106B/206B is directly relevant to the research I have been conducting in the AutoLab. I generate homework problems and labs for the students and grade both. I teach Discussions with small groups of students and hold office hours to reinforce the course material.

Conduct research in Robotic Manipulation, Computer Vision and Path Planning incorporating Machine Learning and Artificial Intelligence. Developed models for Toppling and Pushing to increase access to robust parallel jaw and suction cup grasps. Using a hybrid of Imitation Learning and Model Predictive Control, I significantly reduced the number of trajectories required versus a traditional Imitation Learning path planning policy learned from the Covariant Hamiltonian Optimal Motion Planner (CHOMP). Using Depth Images, I am developing a self-supervised policy to predict Rigid Body Transforms of objects. This policy can then be used to accelerate training of downstream tasks like predicting grasp quality, and can be used to reduce the number of labelled datapoints required for downstream tasks.

I created a Messenger Bot which would allow the developers on the Crowdsourcing team to better understand how Facebook user claims were processed through the Crowdsourcing pipeline for faster development. To do this, I studied the feedback from Facebook Employees outside of the Crowdsourcing team and came up with a roadmap of functionality I deemed to be essential or desirable in the Messenger Bot, and presented this to my team. I then implemented the features according to this roadmap I laid out.

Using ElasticSearch, I modified the Facebook Career website to remove Facebook’s dependence on Salesforce to back the data displayed on each page. This had two benefits: speed and reliability. It reduced certain page load times from upwards of 10-15 seconds to under 1. Facebook was limited in the number of Salesforce DB queries it could make each day. This project reduced the number of times Facebook reached its daily capacity.

Irvine, California, Us

Using ROS, I created an interface to easily send collision free joint trajectory commands to the 7Bot and Yumi robots. This system could use several different path planners, such as CHOMP and RRT* to determine the optimal collision free path. I used the Yumi robot for real time path planning and collision avoidance to play a game of chess. This work was part of the Brass project.

Sunnyvale, Ca, Us

Added strong-authentication to OpenMRS (Open Medical Record System) using StrongKey CryptoEngine to enable encryption, digital signatures and FIDO Universal 2nd Factor (U2F). StrongAuth is in the process of contributing this feature back to OpenMRS.

Taught the equivalent of AP Computer Science

San Jose, Ca, Us

Master Of Science - Ms, Electrical Engineering And Computer Science (Eecs)

Bachelor Of Science - Bs, Electrical Engineering And Computer Science (Eecs)

High School