Sep 2020 — May 2024 · New York, NY
Software Engineer, FounderTherapy
Nearly four years on product tracking and verification software for FounderTherapy's client Lucid Green — a Django platform that turns laboratory test results into scannable, printed product labels, at volume.
Performance work
The part of the job I enjoyed most: instrumenting the system, finding what was actually costing time and money, and fixing it.
10×
faster PDF label generation, with 43% lighter output files
1,840 → 139 ms
median response on the heaviest API: 55.2% → 7.6% of web dyno load, throughput 0.257 → 0.416 RPM
30s → 454 ms
an endpoint that timed out entirely when loading 4,000 records
- Deployed New Relic across the stack for performance monitoring and API availability alerting.
- Profiled critical flows to find bottlenecks, then prioritised the fixes by what they cost the business.
Labels and printing
- Built a custom browser-based product label editor used to design the labels themselves.
- Wrote a PDF label generator producing 65–73 labels per second.
- Supported printing through PDF, BarTender and NiceLabel workflows.
- Generated barcodes and QR codes in PDF, SVG and PNG.
Applications and integrations
- Developed Django web applications and APIs against real user requirements, not abstract specs.
- Built and maintained laboratory integrations that pull test results automatically and place them onto generated labels.
- Integrated MQTT into a browser extension to stream user browser logs and events into BigQuery.
Keeping the lights on
- Kept Django, the Heroku stack and PostgreSQL current and healthy through repeated upgrades.
- Wrote scheduled backups covering databases, file assets and code repositories.
- Wrote a cleanup job that removes orphaned files from S3 buckets and Uploadcare.
Technologies
Django · PostgreSQL · Heroku · AWS S3 · BigQuery · MQTT · New Relic · HTMX · JavaScript · Bootstrap