I build systems that reduce the administrative overhead between a team’s intent and its execution.
CONNORTESSARO
Three independent projects addressing decision capture, inference accounting, and context verification.
Each one began as a question about where a system loses information, and each is built from the interface down to the data layer underneath it.
Ringi
A thread records what a team said. It does not record what the team decided.
Ringi operates as a structured decision instrument inside Slack. On session initiation, the platform messages each stakeholder independently, capturing their position and the constraints underneath it before anyone sees another participant’s answer, then synthesizes those responses into a brief that names the crux, presents the options, and returns a recommendation to the people the decision affects.
Read the case studyEach stakeholder receives a direct message and answers without visibility into the other responses, which captures the constraint underneath a position rather than a reaction to whoever answered first.
Phantom
A streaming request can stop being visible to the client before the system has finished accounting for it.
Phantom operates as an OpenAI-compatible inference gateway with prepaid keys. The platform verifies the key and its available budget before routing a request through Vercel AI Gateway, captures usage while the response streams rather than after it completes, and settles every request in integer micro-USD against a PostgreSQL ledger — including requests the client abandons mid-stream.
Read the case studyThe request arrives against the OpenAI-compatible contract that existing clients already target. The gateway authenticates the prepaid key and confirms available budget before any upstream call is made.
Kizuki
Stored context is only useful for as long as it survives new evidence.
Kizuki operates as a local-first intelligence layer that organizes scattered work activity into a Markdown vault on the user’s own filesystem. The agent returns a structured payload and deterministic JavaScript applies the write, which keeps file mutation separate from model reasoning, preserves handwritten notes outside the managed section, and attaches a source receipt to every signal so any claim can be traced back to the meeting or update that supports it.
Read the case studyA proposed update arrives alongside the vault record the workspace already holds, so the draft is evaluated against existing evidence rather than in isolation.
02 — Professional and open source
Production systems, delivered on teams that were already running.
Software Engineer Co-op
I work on labor and capacity planning, spanning the Snowflake pipelines that produce the forecast, the AWS infrastructure the pipelines run on, and the interfaces that carry the resulting numbers to fulfillment-center teams. Rewriting the labor-planning pipeline reduced its runtime from roughly 60 minutes to 10, measured against its scheduled production runs.
Labor-planning pipeline runtime
Contributions to Prisma ORM
I fixed millisecond DateTime handling in the SQLite driver adapter, where values were losing sub-second precision on the round trip, and added regression coverage so the behavior stays pinned. A second change corrected the distinct-field documentation.
I work on problems where understanding the system is half of the work.
I am a computer science student at Northeastern University with a mathematics minor, currently on co-op as a software engineer at Chewy in Boston. The problems I return to are the ones where information degrades as it moves between people and systems — a disagreement that never gets stated, a billing path that opens when a stream disconnects, a stored assumption that quietly stops being true.
Each of those questions became a project, and each required building across the full stack: the interface, the application state underneath it, and the evidence that the two still agree.
Read the full backgroundSmaller projects, built to answer a single question.
Match data and music, treated as systems worth reading carefully.
LeagueIQ
A match analysis tool that reads a League of Legends game through its decision points — objective contests, rotations, and participation across the timeline — rather than through the final scoreboard.
A score for listening
A Python-generated listening map of Le1f’s “Wut”, rendering rhythm, texture, and structural repetition as a single visual score for a music course.