About
Software engineer in the San Francisco Bay Area. I work on the systems that watch other systems: metrics, logs, traces and alerting, and the search that makes all of it findable.
I am a generalist. Since 2013 I have moved from frontend work to backend and infrastructure: observability, compliance tools, search and data pipelines.
The thread through most of it is search and small languages. I have run Elasticsearch clusters and built the indexing and ranking around them. Twice the job needed a parser for a query language: once to translate metrics queries, and once so that a UI could check Lucene queries as people typed them.
I took an MS in computer science at Santa Clara University while I worked. Outside of work I build plutonium.io, which runs the agents you build.
Work history, as a trace
A trace shows one request as bars on a time axis, one bar for each service it passed through. This one is my career, drawn to scale. Select a row that has a plus sign to read what I did there.
- Jobvite Frontend DeveloperFeb 2013 to Mar 2014, 1 yr 1 mo
- hiQ Labs Software EngineerNov 2014 to Apr 2016, 1 yr 5 mo
-
hiQ Labs Senior Software EngineerApr 2016 to Aug 2017, 1 yr 4 mo
- Architected and built the customer-facing product: the Django backend, the Postgres database with its schema and migrations, the ingest pipelines, and the React frontend.
- MS in Computer Science and Engineering Santa Clara UniversitySep 2016 to Dec 2017, 1 yr 3 mo
-
Lyft Software Engineer (T5, senior)Feb 2018 to Jun 2022, 4 yr 4 mo
Generalist on the Observability team, on cross-team services and libraries for metrics, logging and tracing.
- Led the build of a new alerting service to MVP. Every engineer at Lyft used it, and it evaluated about 30,000 alerts every minute.
- Wrote a parsing library for a metrics query language. It translated tens of thousands of queries at once when Lyft moved to an in-house metrics system, which saved thousands of hours of developer time.
- Worked on the Kubernetes sidecar and DaemonSet containers that collect logs and metrics.
- Maintained a self-managed, federated Elasticsearch cluster of more than 100 nodes for log search.
- Maintained the logging, metrics and tracing libraries for Python and Go.
- Built observability tools, pipelines, services and Grafana dashboards.
-
Bloomberg Senior Software EngineerDec 2022 to Aug 2024, 1 yr 8 mo
Bloomberg Vault Compliance Applications: web apps for compliance officers.
- Wrote a parser for the React UI that validates and parses Lucene queries, and a query editor on top of it with syntax highlighting and validation. Translation runs on the abstract syntax tree the parser produces.
- Built Python Flask API services on SQLAlchemy and upstream data sources.
- Made developer tools: Bash scripts that translate OpenAPI-generated Python client code, a Go CLI that downloads and processes release artifacts, and an Alembic tool that manages migrations in CI.
-
Roblox Senior Software EngineerMay 2025 to now, 1 yr 6 mo
Creator Knowledge team: documentation search, indexing and data pipelines.
- Built a PySpark pipeline and a monthly Airflow task that aggregate millions of search-click and documentation-view events to S3, then fed that engagement data into indexing to boost the results people use. Search click-through rose about 10%.
- Built and deployed a web crawler that extracts metadata and page content, so that pages owned by many separate teams across the Creator Hub Next.js app are searchable.
- Maintain the Elasticsearch cluster, indexing pipelines, document generation, ranking and search UI.
- Built PySpark and Airflow jobs that move offline personalization data into online storage for real-time use in Creator Hub, which powers A/B tests and personalized views.
Search my work
I have worked with Elasticsearch for years, and I once wrote a parser so that a UI could check
and colour Lucene queries. This is a small search of that kind, over the 16 items of
my résumé. It reads Lucene query syntax, and the fields are company, tech and year.
16 items
-
Built a PySpark pipeline and a monthly Airflow task that aggregate millions of search-click and documentation-view events to S3, then fed that engagement data into indexing to boost the results people use. Search click-through rose about 10%.
-
Built and deployed a web crawler that extracts metadata and page content, so that pages owned by many separate teams across the Creator Hub Next.js app are searchable.
-
Maintain the Elasticsearch cluster, indexing pipelines, document generation, ranking and search UI.
-
Built PySpark and Airflow jobs that move offline personalization data into online storage for real-time use in Creator Hub, which powers A/B tests and personalized views.
-
Wrote a parser for the React UI that validates and parses Lucene queries, and a query editor on top of it with syntax highlighting and validation. Translation runs on the abstract syntax tree the parser produces.
-
Built Python Flask API services on SQLAlchemy and upstream data sources.
-
Made developer tools: Bash scripts that translate OpenAPI-generated Python client code, a Go CLI that downloads and processes release artifacts, and an Alembic tool that manages migrations in CI.
-
Led the build of a new alerting service to MVP. Every engineer at Lyft used it, and it evaluated about 30,000 alerts every minute.
-
Wrote a parsing library for a metrics query language. It translated tens of thousands of queries at once when Lyft moved to an in-house metrics system, which saved thousands of hours of developer time.
-
Worked on the Kubernetes sidecar and DaemonSet containers that collect logs and metrics.
-
Maintained a self-managed, federated Elasticsearch cluster of more than 100 nodes for log search.
-
Maintained the logging, metrics and tracing libraries for Python and Go.
-
Built observability tools, pipelines, services and Grafana dashboards.
-
Architected and built the customer-facing product: the Django backend, the Postgres database with its schema and migrations, the ingest pipelines, and the React frontend.
-
Software Engineer.
-
Frontend Developer.
This chart is your visit
Most of my work is on systems that watch other systems. This is a small one: the page measures four signals in your browser and checks an alert rule against each one, twenty times a second. Nothing is sent anywhere.
- pointer speed > 1500 px/s for 1 s ok
- scroll speed > 2500 px/s for 0.5 s ok
- frame time > 34 ms for 2 s ok
- idle time > 20 s ok
Contact
For work, or about anything on this site, write to me.