Founded Xavi Group to build at the intersection of human and machine intelligence. Two flagship projects anchor the work.

Xavi is an agentic harness that gives AI systems durable continuity and a real body of tools across a live codebase — persistent memory, a knowledge graph, and infrastructure for collaborative intelligence that spans thousands of sessions.

blackbear is a private, local-first, end-to-end-encrypted everything-app — notes, documents, calendar, tasks, spreadsheets, messaging with voice and video, audio, and more — rivaling the Proton, Google, and Microsoft suites while keeping your data yours. It runs across web, native desktop, CLI, and mobile, built on a single Rust crypto substrate and over a million lines of code.

Lumin was acquired by ABB in January of 2025. My first project post-acquisition was an integrations cloud service that builds on concepts developed three years prior, allowing for secure and stable connections with third party device APIs by acting as a connection proxy for our IoT devices. Notably, this enabled the rollout of additional cloud integrations without requiring any changes or software deployments on IoT devices.

Building on the first project, my second project was writing a connection adapter leveraging the authorization and device APIs of our hub product to enable multi-hub systems. Multi-hub systems leverage the local network and mDNS to permit one software system to control and monitor circuits from up to four local hub hardware devices (over 48 single poll circuits).

I was promoted to Senior Software Engineer as we started working on a transition to a new software stack. My first project was creating the scaffold for our new hub software application, and an internal service within it for managing connections to the hub hardware and eventually third party devices. During this time I also developed wrappers around logging and tracing packages for increased observability and debugging of backend software code.

In the process of bringing in NATS, due to my past experiences with Kafka I created my second internal library, a library that serves as an abstraction over eventing. The library now has adapters for around four different streaming platforms, allowing us to use the same logic for writing and reading events locally as we do in many cloud services.

After working on the skeleton of the hub software application, I worked on standing up a number of core micro services in the cloud. The most important being a fine-grained authorization model (third internal library) and authorization service, featuring a mechanism that generates test cases based on the authorization schema.

Additionally, I led the development of a new micro service for managing device software, built on the concepts used for the internal CLI / API written a few years back.

I joined Lumin as its 4th engineer. In my first few months I worked with our devops engineer and built a data pipeline on top of Apache Kafka. I implemented an adaptable, horizontally scalable Golang micro service that now handles over five billion requests per day, collecting readings from IoT devices across the country. Over the past four years, the service has scaled and operated with a 99.9999% uptime.

I spent time reviewing time-series data stores and worked with the engineering team to transition from OpenTSDB to Timescale Cloud. The switch improved queries for our in-app data browser across the board, and in some instances by an order of magnitude. The service also led to a sharp reduction in data storage costs when combined with the adoption of Apache Iceberg.

I also worked on software services on our IoT device (the Lumin LSP), creating an internal CLI / HTTP API for managing device software deployments. We've used that tool for device deployments for the past two/three years.

During this brief internship I worked with the Decision Engine team on an ongoing project to refactor core application functionality from Ruby on Rails to Java Spring. I made modifications and additions to controllers, models, and view layers in both frameworks. Additionally, I worked to create Internal Admin pages for new features with React, and modified some Python project configuration scripts.

Over the course of eight months I worked with numerous agile teams to develop production Ruby on Rails API features, frontend React features, and proof-of-concept projects. I also improved code quality in the Rails codebase by enforcing stricter linting and refactoring areas of legacy code.

The most interesting proof-of-concept project was the creation of a prototype chatbot, which showcased the potential features we could provide through integrations with Google Home and other IoT devices. I worked directly with the VP of Engineering and the Chief Medical Officer. I presented this project internally at one of the "All Hands" meetings, and it was presented at health tech conferences by the company leadership team.

In my second time as a TA for this course I focused on improving the feedback I gave to students on their projects and providing a better experience for the students who came to my office hours. I also took on additional responsibilities in leading exam review sessions and proctoring exam sessions.

My first time as a TA was a great learning experience. I graded projects and gained valuable experience with both reading code and providing constructive feedback. I helped with exam review sessions and created study materials to help prepare students for the exams.

Working with Scout Studio (a Northeastern student-led design studio) taught me a lot about user experience design and the design process. We worked with a Northeastern startup to create a brand for an inclusive suncare product, which culminated in the development of a landing page and packaging designs. I developed the landing page using React and Prismic (CMS).

At CoS I optimized API tests across the entire application and reduced the test suite run time by over 30%. I also developed a few features for the Open Science Framework platform, built using Django and Python. I presented at PyOhio some of the strategies used for transitioning test code from Nose to PyTest.

My friend Tyler Cosgrove and I became the youngest members accepted into the University of Virginia iLab program after we pitched our idea for a social shopping platform with the ability to recommend more accurate clothing sizes. For nearly a year we worked on the platform, which we developed using Meteor and MongoDB in addition to a web scraping component built with Python.

From my first year in university through my last, I challenged myself to take the courses where I would be pushed to learn the most. Having been a self-taught developer before entering college, I was amazed by how much I didn't know, and thrilled to be able to put names to some of the practices and patterns I had been using for years.

The second part of my degree was in business and entrepreneurship, which carried over from my interests in high school. After studying more philosophical and spiritual movements I tailored my entrepreneurial interests into the formation of socially responsible businesses and social enterprises.

Throughout high school I was interested in computer science and entrepreneurship, and started the first computer science and entrepreneurship clubs at Albemarle in my first three years with the help of mentors and faculty. Being in close proximity to the University of Virginia, we were able to tap into the computer science and entrepreneurship programs at the university to provide guest lectures and networks for students.