Michael Zavarella

Staff software engineer with 11 years architecting distributed, multi-tenant systems, and the production AI/LLM platforms now running on top of them. My work spans low-latency trading infrastructure, event-driven platform engineering, observability at scale, and LLM-backed products.

Lately I build the orchestration, tooling, and observability layers that make LLM applications reliable in production: tool and function calling, response streaming, and tracing for cost and latency. Along the way I've cut cloud spend 30%, lifted uptime from 99.5% to 99.99% for a 50+ engineer org, and shipped GraphQL platform layers, API gateways, and async processing infrastructure that improve velocity across the stack.

Experience

1. Staff Engineer — Smarter Technologies
   Remote · Dec 2025 – Present

   Lead the architecture of a distributed, multi-tenant insurance payer interaction system — the integration backbone connecting EDI, eFax, RPA, voice AI, and human-in-the-loop workflows — including the async orchestration and observability layer behind an AI voice-calling platform on AWS/EKS. Directing a 3-engineer team.

2. Staff Engineer — McGraw Hill
   Remote · Aug 2025 – Dec 2025

   Led development of an LLM-backed learning assistant adopted by 50,000+ students, shipping v1 in ~2 weeks. Built tool/function-calling integrations for per-student personalization and stood up LLM observability — tracing, cost/latency monitoring, and production dashboards — with Langfuse.

3. Senior Engineer — Capital One
   New York, NY (Hybrid) · Oct 2024 – Aug 2025

   Technical lead and architect for a 5-person observability platform team. Founded the Engineering Excellence Architecture & Design Review Board, drove release frequency from biweekly to daily, and introduced GraphQL across the org to decouple data access from UI.

4. Staff Engineer & Engineering Manager — Yummly
   San Carlos, CA (Remote) · Aug 2021 – Mar 2024

   Built and led a 5-engineer platform team owning CI/CD, AWS, backend systems, and internal tooling. Cut infrastructure cost 30% and lifted uptime from 99.5% to 99.99%, led US/Brazil internationalization (+20% subscribers), and powered DuckDuckGo's recipe widget.

5. Lead Engineer, Infrastructure & Platform — DV Trading
   New York, NY (Hybrid) · May 2019 – Aug 2021

   Built a low-latency, highly durable tick-level market-data store feeding backtesting and replay for traders in Chicago, NYC, and London, plus async reporting for trading-desk reports. Designed an HA CI/CD system (Jenkins + Docker Swarm) across three global offices.

6. Software Engineer — Capital One · Bioclinica · Blue Cross Blue Shield
   NY · PA · May 2015 – May 2019

   Backend lead on knowledge-management and search at Capital One, lead customer-integrations engineer at Bioclinica, and lead engineer on an internal marketing research platform at Blue Cross Blue Shield.

Tools & Tech

Education

Widener University — B.S. Mathematics · 2011 – 2015

Built

• Clojure.org Documentation

  Authored official guides on higher-order functions and destructuring, and contributed bug fixes to the Clojure core libraries.

  Higher Order Functions → Destructuring →

• GraphQL.org Documentation

  Contributed documentation covering Clojure library representation.

  GraphQL.org →

Elsewhere

github · linkedin · contact@zavarella.me