About
Backend-first full-stack engineer in New York — I make data paths fast, platforms maintainable, and write up the trade-offs so the next engineer doesn't have to reverse-engineer the decision.
APIs, data stores, and the systems around them — subscription platforms, payment rails, market data.
21s → 250ms
Core API latency
40%
Off API p95 latency
99.99%
Pipeline uptime
Case study
21s → <2s → ~250ms
A Data-as-a-Service API was running BigQuery DML on the hot path — 20–30 second queries under load. Storage Write API over gRPC got responses under 2 seconds. Moving the read path to PostgreSQL + EF Core brought hot queries down to roughly 200–300ms. Three architectural decisions, not query tuning.
I'm a full-stack engineer who has always been wired to understand how things work — take them apart, learn the mechanics, put them back together better. That instinct turned into programming in 2015 and grew into a career building tools that solve real problems at scale.
I'm strongest on the backend — schema design, migrations, gRPC, message pipelines — but I routinely own the full path when it matters: Next.js frontends, Auth0, multi-pod caching, microfrontend shells. The pattern is the same: find the bottleneck, pick the right tool, ship with tests and measurable latency.
At Visa I worked on payment and eligibility systems at production scale (20M+ monthly requests on the eligibility path) — modernizing .NET and Spring Boot services, replacing hundreds of stored procedures with APIs, and building the first real-time benefit-redemption pipeline between VCES and VDBP at 99.99% uptime.
After Visa I spent a self-directed stretch shipping mobile apps, a CRDT collaborative editor, and RAG prototypes — the kind of end-to-end work that keeps API design honest.
At S&P Global I started on Data-as-a-Service — taking core API latency from ~21 seconds to under 2 with the Storage Write API, then to roughly 200–300ms on PostgreSQL, while modernizing the frontend and migrating dozens of routes off BigQuery. That expanded into Data Studio: a microfrontend platform (Module Federation, Fastify BFFs, shared contracts) serving DaaS, mData, and other remotes.
This site is the latest iteration of that habit — portfolio v2 on Next.js 16 with an MDX essay pipeline, static generation, and the same supply-chain discipline I document in npm audit triage.
BigQuery Storage Write API over gRPC, then PostgreSQL + EF Core for OLTP — two architectural moves that took an interactive API from 21-second tails to ~200–300ms.
From last-write-wins string sync to Yjs binary updates, append-only Postgres persistence, and object-storage snapshots — how a production-style collaborative editor handles concurrency, reconnects, and durability.
Expo Router, Firebase split-brain messaging, Zod as schema source of truth, why I skipped an API gateway, and what a frozen social-mobile codebase taught me about shipping IRL-first software.
Twenty technical essays across enterprise platform work, mobile archives, and side projects — with decision tables and measured outcomes. Browse all writing →
October 2025 — Present/New York, NY
S&P Global Inc.
February 2025 — October 2025/Remote
Self-Directed
September 2022 — February 2025/Miami, FL
Visa, Inc.
June 2022 — August 2022/Boston, MA
Wayfair, Inc.
September 2020 — July 2021/New York, NY (Remote)
Digital memorial platform (remote)
Faster queries, lower latency, leaner builds. A 21s API path that lands at ~250ms changes what a product can do — that's the work I optimize for.
When a system moves money or market data for millions of users, correctness and uptime aren't optional. I design for failure modes, not just the happy path.
Every engineering decision should move a number — p95 latency, error rate, coverage. I orient around outcomes, not activity.
Readable code and clear writing are the same discipline. The next engineer — often future me — should understand the why without an archaeology dig.
September 2017 — June 2022
Williams College
Next
Compare notes on something you're building
Performance bottlenecks, realtime sync, payment rails, or how you triage a messy dependency tree — I read every message.