Support engineer moving into platform and data engineering. Four years resolving production incidents in SQL Server, cloud networking and endpoint security for critical-infrastructure customers, backed by a self-hosted Linux stack I design and run outside of work.
Case study · DotHub
The first Docker build of DotHub, my job-search app (featured below), failed at random, and a retry usually passed. Retrying until it goes green is the easy answer. The error message pointed at something real.
Some builds stopped with SQLITE_BUSY: database is locked while collecting page data. It depended on timing, so it looked like noise.
The build imports every route, so 15 workers opened the same fresh SQLite file at once and raced to switch it to WAL mode. The losers failed, even with a 5-second busy timeout.
During the build, each worker gets its own empty, migrated in-memory database. The build has no data to read, so there's nothing to share.
15 → 0
Workers sharing a database file during the build, so the lock can't happen. A 9-line change.
SQLite's backup API wrote each copy in WAL mode, and the read-only connection that checked it couldn't clean up its -wal and -shm files on close. Every day left an orphaned pair that pruning never removed. The check now switches the copy to a rollback journal first, and old leftovers get swept, with tests.
1 self-contained file per backup
Thousands of requests with forced garbage collection and heap-snapshot diffs, plus 30 laps of in-browser navigation, found no memory leak. They did find a validation library riding into the browser, a Markdown renderer loading on every page, and a rate limiter that grew with every one-off visitor.
442 → 371 KB of JavaScript on the main page
Featured project
A scheduled bot scores new postings; the app tracks applications, follow-ups, interviews and offers. I built it end to end and run it in production on my own Linux host.
A versioned, append-only data store for a career, where every resume is a selection over the data, not a copy. JSON Schema contracts define every shape, and an audit flags any number in the prose that isn't in the bullet's metrics.
A PowerShell tool that scans up to 25 Carbon Black App Control logs at once and pulls out the most repeated errors, their codes, timestamps and counts, plus any IP addresses. It writes 4 structured files, including a master diagnosis summary.
11 services from one Docker Compose file, non-root containers and env-file secrets, with Cloudflare Tunnel as the only ingress. I migrated one service's PostgreSQL to a VectorChord build for vector search, and pinned images to major tags after an unpinned pull broke it.
Experience
Carbon Black Cloud, App Control and EDR for US critical-infrastructure and public-sector customers running 100 to 5,000 endpoints. Cleared a 53.5M-file backup backlog in one maintenance window by tracing it to missing SQL Server permissions, not the index corruption first suspected. I own 10–15 cases a month end to end, including Severity 1 incidents handed off across time zones, and diagnose sensor connectivity failures across TLS inspection and certificate trust with curl and packet-level evidence.
Resolved Azure API Management and web app failures by reading platform logs and tracing request paths through NSGs, firewalls and gateway policy. Used PowerShell, Bash and SQL daily, and trained teammates on Azure fundamentals and log reading.
Tested autonomous air-quality robots in client buildings, analysing device logs and escalating reproducible defects with evidence. Wrote a Bash script that automated the daily test-rig setup.
Four-month immersive certificate: cryptography, threat intelligence, networking, systems administration and system security.
Contact
I'm based in Colorado Springs and looking for remote roles in the US. Email is the fastest way to reach me.