Hello — I'm Ada.
I build the systems your product quietly depends on.
Eight years across payments infrastructure, developer platforms, and data pipelines. I care about the parts nobody demos: correctness under load, migrations that do not lose rows, and on-call rotations that let people sleep.
Available for contract work from March.
- Years shipping backend systems
- 8
- Daily requests served at peak
- 40M+
- Availability on my last platform
- 99.98%
Lisbon, Portugal
Selected work
A few projects I can describe publicly. Happy to go into detail on any of them.
Ledger migration
Moved a double-entry ledger holding nine years of transactions from a single Postgres instance to a sharded setup, with no downtime and no reconciliation drift.
Cut p99 write latency from 840ms to 62ms.
- Postgres
- Go
- Kafka
- Terraform
Deployment platform
An internal platform that gave 60 engineers self-service deploys with automatic rollback, replacing a process that needed a platform engineer in the room.
Deploys went from 4/week to 40/day.
- Kubernetes
- Go
- ArgoCD
- OpenTelemetry
Rate limiter
A distributed token-bucket limiter that degrades gracefully when its coordination layer is unavailable, rather than failing open or shut.
Absorbed a 12× traffic spike without shedding valid requests.
- Rust
- Redis
Schema registry
Compatibility checking for event schemas in CI, so a breaking change fails the build instead of a consumer at 3am.
Eliminated a recurring class of production incident.
- TypeScript
- Protobuf
- GitHub Actions
About
I started in backend work because I liked problems where the answer was either right or wrong. I stayed because the interesting version of that question — right under what conditions, wrong how often, and what happens next — turns out to be most of engineering.
Most of my work is on systems that handle money or identity, where being approximately correct is the same as being broken. That constraint shapes how I build: small reversible steps, real observability before scale, and a strong preference for boring technology that has already failed in public.
I write occasionally about distributed systems and the operational side of software. I mentor two junior engineers through a local programme, which has made me a considerably better explainer.
Languages
- Go
- Rust
- TypeScript
- Python
- SQL
Infrastructure
- Kubernetes
- Terraform
- AWS
- Postgres
- Kafka
Practice
- Distributed systems
- Observability
- Incident response
- Technical writing
Experience
Staff Engineer at Northwind Payments
Technical lead for the ledger and settlement systems processing payments across fourteen markets.
- Led the sharding migration described above with zero customer-visible downtime
- Introduced a reconciliation harness that catches drift within one hour rather than at month-end
- Reduced on-call pages by 71% by fixing causes rather than adding alerts
Senior Platform Engineer at Basalt
Built the internal developer platform from an empty repository to the tool every team deployed through.
- Designed the deploy pipeline and rollback mechanism used by 60 engineers
- Cut median time-to-first-deploy for new hires from nine days to four hours
- Ran the migration off a monolith into services that could be owned independently
Backend Engineer at Kestrel Data
Data ingestion and processing for a analytics product handling several billion events a month.
- Rewrote the ingestion path in Go, cutting infrastructure cost by 40%
- Built the schema registry that later became an internal standard
Writing
Occasional notes on distributed systems and the operational side of software.
Your retry logic is a denial-of-service attack
Exponential backoff without jitter synchronises your clients. A walkthrough of how a well-intentioned retry policy amplified a minor blip into a four-hour outage.
Migrations that cannot lose data
Dual writes, backfills, and the verification step most guides skip. What it actually takes to move a table nobody is allowed to notice moving.
On-call is a design problem
If your rotation is painful, the fix is usually upstream of the rotation. Notes from cutting our page volume by two thirds without silencing anything.
Let's talk
I'm open to contract engagements and the occasional advisory conversation. If you have a system that is creaking under load or a migration nobody wants to own, that is usually my kind of problem.
hello@example.comI reply to everything within two working days.