Personal siteAda Okonkwo
A career that reads like a career
An editorial hero, a timeline of roles with a real open end, selected work that opens in place, and a résumé action that prints the page instead of promising a file that was never bundled.
- Built from
- ProfileHeroSocialLinksExperienceTimelineProjectShowcaseBadgeButtonSeparator
- What the browser test drives
- Open and close a project, follow the related link, read the timeline, and press Print a CV to see the print layout strip its own controls.
- Appearance
- Cobalt with Inter at the layered default — the same palette as the marketing page, and a completely different page.
Preview
Available for consulting from January
Ada Okonkwo
Distributed systems engineer
I build storage layers that keep working when a region does not. Twelve years across two continents, most of it on the part of a system that is only interesting at 3am.
- Based in
- Lisbon
- Since
- 2014
- Focus
- Replication, consensus
Where the work has been
Three roles, each one a longer problem than the last. The highlights are the things I would still defend in a review.
Staff engineer, storage platform
PresentNorthwindRemote, Lisbon
Own the replication layer that keeps four regions consistent. Two engineers, one on-call rotation, no database vendor.- Cut p99 write latency from 180 ms to 74 ms by moving quorum acknowledgement off the request path
- Ran the migration of 2.4 PB to a new log format with no read outage
- Raft
- Rust
- Postgres
- On-call
Senior backend engineer
ContosoBerlin
Grew the ingestion pipeline from 40 to 9,000 events a second and, more usefully, wrote down why it had stopped scaling twice.- Replaced a queue-then-batch worker with a backpressured stream: 60 % less memory, same throughput
- Introduced the deploy-attached alert that halved mean time to identification
- Go
- Kafka
- ClickHouse
Backend engineer
Lumen AnalyticsLisbon
First infrastructure hire. Everything from the CI pipeline to the on-call document that nobody had written yet.- Moved a monthly batch report from 9 hours to 40 minutes
- Python
- Airflow
- Terraform
Selected work
Choose a project to read the reasoning behind it. The grid stays on screen so the work can be compared rather than replaced.
Interactive, not live: every action below changes local state in this page. No network request is made, no market data is fetched and no message is sent anywhere.