Professional

Meta (2025)

Reality Labs

Currently building out a VR & MR runtime that empowers developers to build engaging, immersive experiences.

Sayari (2022-2024)

Sayari Map

Automated supplier mapping and risk screening tool that allows customers to visualize their supply chains, while identifying risk multiple tiers upstream of their known suppliers. This enables screening thousands of these suppliers at a time for many different risks of various severities, and de-noising results with AI-enhanced product categorization.

This involved quite a bit of graph traversal algorithm work - supplier networks quickly grow into tens of thousands of entities as you expand just a few hops upstream, and graphs with hundreds of millions of nodes and edges meant the search space itself could easily exceed a hundred thousand nodes.

The standard breadth-first search approach doesn't end up solving for these problems very well (i.e. visiting the same node as part of separate paths is actually valuable here, depending on how otherwise unique the paths are), so I ended up leading some fairly interesting work around building and optimizing in-memory traversals, along with adding interdependent filters that evaluated combinations of edges along paths to determine validity (relevance) during expansion.

Another part of this effort was R&D related to the correlation between centrality (in the graph theory/network analysis sense) and potential supply chain bottlenecks, which are inherently risky for customers downstream of them.

Sayari Graph

Commercial risk intelligence platform with huge global datasets describing complex ownership and trade networks between corporate entities in a combination of NoSQL, graph databases, and text-optimized (inverted index) data stores.

A majority of my work for this was writing complex queries and aggregations with Elasticsearch, as well as leading development of a new set of API endpoints focused on trade data, enabling customers (and our own UI) to surface key data points with user-friendly parameters.

API design complexity tends to scale with the comprehensiveness of data models and access patterns involved, which were both extremely dense in this case. So, part of my effort was maintaining ease-of-use and cohesiveness of endpoints (abstractions and denormalization vs. limited projection of relevant fields), while avoiding inefficiencies like N+1 query problems/query amplification. There were also a lot of NoSQL and 'shortest path to X' problems to solve here, too, where X could be any combination of distinguishing information about an entity or its relationships to other entities.

Amazon Web Services (2020-2022)

Providing Secure Software Project Development Environments

Patent: US-11838296-B1

Better organization-level control over projects tracked by version control systems, allowing remote governance of what and where code can be pushed from local and cloud-based IDEs, among other configurable policies.

This was the first time that I had worked on an executable client that was actually deployed on a large number of users' machines, as opposed to previous work that was mostly confined to server-side and web apps. Writing a client in Go that interacted with Git actions, made requests on behalf of the user to determine if target repositories are publicly accessible, and acted as a `launchd` or `systemd` daemon, was interesting, along with all of the systems design in implementing it within the distributed cloud services layer.

Ford Motor Company (2017-2020)

Canopy (before public launch)

A security system for pickup beds, AI-powered intrusion monitoring, cloud services-based event handling, and a mobile app to view events and manage/stream from devices.

My initial deep dive into cloud services (AWS), beyond just hosting containerized apps on compute resources and DBaaS; originally developed on top of services like Kinesis for video streaming, IoT Core for communications, SQS/SNS for event-driven workflows, etc.

Communicating with custom hardware over cellular networks was a fun challenge - devices being offline or in locations with bad connections presents problems to solve both in code and system designs. This was also my first experience with React Native for mobile app development, which was a surprisingly easy switch from React JS at the time.

Personal

My GitHub has a few small hobby projects that I've worked on in the past, touching on concepts that I was interested in at the time like concurrency and tokenization.