Six years of backend and data platform work, newest first.

Senior Backend Engineer, Data

I am the primary engineer behind Copilot, Reserv's production large language model (LLM) agent for claims adjusters, working across the Django and LangGraph agent, the Rails backend, the React frontend, and the Node runtime. I built its suite of about 20 tools with human-in-the-loop approval, its conversation and memory engine with automatic compaction, and a safety layer that validates every AI-generated email and note for accuracy, tone, and compliance before it is sent, with an audit record of every tool call. Copilot doubled claims accuracy and gives adjusters back 40+ hours a month.

In 2026 I rebuilt Copilot's knowledge retrieval as an agentic tool loop, which raised retrieval accuracy by 64.3 percentage points on de-identified production fixtures and shipped behind a feature-flag ramp with a kill switch. I ran the coordinated four-repository cutover to the AG-UI agent streaming protocol and root-caused the production memory failures and approval double-fires it exposed. I also built the company's generative AI evaluation framework and LLM observability from scratch, including a shared de-identified benchmark dataset, vendor adapters for LangSmith and Arize, and the Datadog LLM observability tracing that informed a six-figure annual vendor decision.

On the data side, I built and then replaced the claim-document ingestion pipeline: an event-driven Kafka consumer with idempotency receipts and dead-letter replay that let us delete the legacy AWS Textract integration. I own the pipeline that produces the bordereaux and loss-run reports that about 20 insurance carriers require, cut dbt model runtimes by 28%, reduced Docker image sizes by up to 76%, and built most of the team's Datadog alerting and the queue-depth-driven AWS Elastic Container Service (ECS) autoscaling that replaced manual capacity checks.

Senior Backend Engineer, Platform

Merit runs on Go microservices. On the platform team I took the core services (authentication, notifications, data ingestion, graph management) from first deployment to more than 250,000 users. I expanded integration test coverage by 60% across every core service so that cross-service failures surfaced in the test suite instead of in production, built the integration framework that connects Snowflake and Fivetran to the platform and syncs more than 3 million records a day, and rebuilt the notifications service from a prototype into production infrastructure that can send millions of messages a day.

Backend Engineer, then Senior Backend Engineer, Platform

I joined Neutron as a backend engineer and left as the senior engineer on the platform. In the first year I led the move of the legacy monolith into Docker, shipped over 10 features for the software as a service (SaaS) platform, including the matching algorithm behind over 80% of company revenue, and wrote the unit and integration suites that took the system past 80% code coverage. In the second year I built the GitHub Actions continuous integration and delivery (CI/CD) pipeline from scratch, cutting release cycle times by 75%, moved the platform onto ECS, and replaced manual AWS console management with Terraform. I also recruited and onboarded two junior engineers.