OpenTeams Enterprise Solution Architect started with a small team of one full-time engineer and two part-time engineers working one day a week. The part-time engineers both had extensive prior experience in the probabilistic programming space. Over the last year, as a result of helping Meta achieve their goals, the ESA grew the team to allow the part-time engineers to move to half-time. The scope also increased beyond writing documentation and tutorials to include adding major software features.

OpenTeams’ Enterprise Solution Architects installed a data science platform (Nebari) that provided shared, scalable compute, and storage across all teams. They created a pipelining solution to demonstrate best practices in MLOps for all of their teams to adopt, including automated data ingest and validation, model training, and dashboards to reduce errors, provide monitoring, and make model experimentation easy.

OpenTeams Enterprise Solution Architect updated their risk model to run on Python and utilize fast numerical libraries like NumPy and Numba. The ESA also moved the model to a high-performance deployment on computation clusters using PySpark.

OpenTeams’ Enterprise Solution Architects introduced machine learning capability to the simulation by building a Python API and an MLOps pipeline that built reinforcement learning models. Reinforcement learning models exhibit autonomous behavior that allow simulation entities to adapt to complex situations, even if the human operator has not anticipated the situations. These models are exploratory and often find novel strategies that the human operator did not imagine.

OpenTeams Enterprise Solution Architect built a data pipeline to connect the machines and CRMs. They centralized the data and built dashboards and tools to help the engineers monitor and control the assembly line.

OpenTeams Enterprise Solution Architect was hired by Meta to lead a small team of three part-time engineers who had experience on similar projects, such as NumPy and SciPy. Over the last three years, as a result of achieving Metas goals, the ESA grew the team to 15+ people dedicated to PyTorch development.

OpenTeams Enterprise Solution Architect built an open source, cloud-based data science platform to allow them to scale their work, including a stable, reliable, automated workflow to preprocess new data each day, algorithms to feed various data sources into a single dataset, and algorithms to analyze the data according to their customers’ needs.

OpenTeams Enterprise Solution Architect converted the model to a Python package with an automated workflow that included a test suite, data validation, and report generation. The ESA and his team built a web-based platform with authentication and data input framework. Additionally, the team built tools that allowed technicians to draw on images of the building architectural drawings as part of the data input which allowed them to easily go from room to room. These were also used for automatic report generation.