const me = {
name: 'Thomas Pedbereznak',
role: 'Senior Software Engineer',
location: 'Seattle, WA',
education: 'BS Computer Science — Northeastern (magna cum laude)',
focus: ['data platforms', 'developer experience'],
status: 'open to new roles'
}me.summary()
Senior software engineer with 5+ years in enterprise software, focused on data platforms and developer experience. I helped deliver a seven-figure data-lake migration, shipped an internal tool used by every tax expert on the team, and tuned Spark pipelines from hours to minutes — balancing hands-on engineering with mentorship and clear stakeholder communication.
me.skills()
Programming
Big Data & Cloud
Testing & DevOps
Web & Full-Stack
me.experience()
Full history in my résumé ↓ — below is a summary of key points.
Migrated six ETL pipelines from Vertica → S3 with full data parity and zero downtime, and rebuilt the CPA Tool onto the lake (the last system blocking Vertica's retirement), part of a seven-figure cost reduction.
Owned 5+ Spark/EMR pipelines and built a TurboTax abandonment pipeline surfacing user drop-off.
Shipped the CPA Tool (Electron) to all tax experts with zero P0/P1s, and built Testable for secure test data.
Enhanced DBDeploy with dataset-level deployments and standardized CI/CD across projects.
Mentored engineers and drove Spark performance wins, cutting job runtimes from hours to minutes.
me.relatedWork()
- Designed and shipped a full online storefront for a handmade knit & crochet maker — React 19 + Vite front end on a Cloudflare Workers backend, live on a custom domain at gnatsknots.com.
- Used Stripe as both payment processor and product database: the catalog reads live Stripe Products, checkout re-verifies stock server-side, and a signed webhook decrements inventory so items flag sold-out automatically.
- Owned the whole delivery path solo — brand and UI, cart logic, a custom-order inquiry flow, security headers on every response, unit tests over the cart and worker logic, and deploys to production.
- Built a data lakehouse for Deadlock (a competitive team game) using Scala, Spark, and Apache Iceberg on AWS S3/Glue.
- Designed typed ETL pipelines with incremental upserts, partitioning, and watermark tracking for reliable reprocessing via Iceberg metadata.
- Optimized performance through Spark tuning, partition pruning, and broadcast joins.
- Managed infrastructure as code with Terraform (S3 buckets, IAM roles, Glue jobs), enabling hero- and item-level dashboards from match-level facts.
- Real-time display of key value indicators for a competitive game, updating live as matches unfold.
- Live outcome / win-probability predictions surfaced alongside the KVIs.
- Built on the Deadlake data platform, turning match-level facts into an at-a-glance realtime view.