Skip to content
Amara Okonkwo-Silva
AMSGet in touch
(Portfolio)Amsterdam, NetherlandsNot looking right now

AmaraOkonkwo-Silva

Staff engineer working on distributed databases, developer tooling and the occasional programming language.

Twelve years of building infrastructure that other engineers build on. I've written storage engines, query planners, CI systems and more internal CLIs than I can count. Currently focused on consensus protocols and observability for multi-region Postgres.

Scroll
  • Rust
  • Go
  • TypeScript
  • Python
  • Kotlin
  • C++
  • SQL
  • Elixir
  • Zig
  • Bash
  • Postgres
  • SQLite
  • Redis
  • ClickHouse
01

About

I started programming by modding games and never really stopped treating software as something to take apart. Most of my career has been spent one layer below where the product lives: storage, networking, build systems, the things that only get noticed when they break.

At my last two companies I led small teams (four to seven people) that owned the database platform. That meant on-call, capacity planning and migrations, but also a lot of writing: design docs, postmortems and the kind of internal documentation that saves a hundred hours a quarter.

Outside work I maintain a handful of open-source libraries, mostly in Rust and Go, and write about the parts of distributed systems that textbooks skip — retries, clocks, partial failures and what happens to your p99 when one disk gets slow.

I'm happiest with a hard, well-defined problem, a whiteboard, and a team that argues kindly.

Based in
Amsterdam, Netherlands
Currently
Staff Engineer, Database Platform at Northwind Data
Languages
Rust, Go, TypeScript, Python
02

Selected work(24)

  1. (01)

    raft-lite

    A minimal Raft implementation with a deterministic simulation test harness.

    Role
    Author
    Year
    2025
    Stack
    Rust, Python, gRPC

    Long description that explains what the project does, why it exists and which trade-offs were made. It covers the design, the failure modes considered and what was surprisingly hard: getting the edge cases right under concurrency, keeping the public API small, and writing tests that actually catch regressions instead of just asserting the happy path.

    • Handles 120k requests per second on a single node in benchmarks
    Live site Source 4.2K

  2. (02)

    pgshard

    Transparent sharding proxy for Postgres with online resharding.

    Role
    Maintainer
    Year
    2024
    Stack
    Go, Kotlin, gRPC, Postgres

    Long description that explains what the project does, why it exists and which trade-offs were made. It covers the design, the failure modes considered and what was surprisingly hard: getting the edge cases right under concurrency, keeping the public API small, and writing tests that actually catch regressions instead of just asserting the happy path.

    • Handles 120k requests per second on a single node in benchmarks
    • Property-based tests with deterministic replay of failures
    Source 4K

  3. (03)

    edge-cache

    An HTTP cache for edge workers with stale-while-revalidate semantics.

    Role
    Maintainer
    Year
    2023
    Stack
    TypeScript, C++, gRPC, Postgres, Docker

    Long description that explains what the project does, why it exists and which trade-offs were made. It covers the design, the failure modes considered and what was surprisingly hard: getting the edge cases right under concurrency, keeping the public API small, and writing tests that actually catch regressions instead of just asserting the happy path.

    • Handles 120k requests per second on a single node in benchmarks
    • Property-based tests with deterministic replay of failures
    • Used in production by three companies
    Source 3.9K

  4. (04)

    tracekit

    Distributed tracing toolkit that keeps overhead under one percent.

    Role
    Author
    Year
    2022
    Stack
    Python, Elixir, gRPC, Postgres, Docker, Kubernetes

    Long description that explains what the project does, why it exists and which trade-offs were made. It covers the design, the failure modes considered and what was surprisingly hard: getting the edge cases right under concurrency, keeping the public API small, and writing tests that actually catch regressions instead of just asserting the happy path.

    • Handles 120k requests per second on a single node in benchmarks
    Source 3.7K

Also built

  1. 05wasm-sandboxRuns untrusted WebAssembly with strict CPU and memory budgets.Kotlin · Zig2021
  2. 06quillA Markdown editor that stays fast on 50k-line documents.C++ · Rust2020
  3. 07orbit-uiA minimal Raft implementation with a deterministic simulation test harness.Elixir · Go2019
  4. 08ledgerTransparent sharding proxy for Postgres with online resharding.Zig · TypeScript2018
  5. 09nimbusAn HTTP cache for edge workers with stale-while-revalidate semantics.Rust · Python2017
  6. 10fastqDistributed tracing toolkit that keeps overhead under one percent.Go · Kotlin2025
  7. 11schema-diffRuns untrusted WebAssembly with strict CPU and memory budgets.TypeScript · C++2024
  8. 12tinyvmA Markdown editor that stays fast on 50k-line documents.Python · Elixir2023
Show 12 moreShow less
  1. 13hyperlogA minimal Raft implementation with a deterministic simulation test harness.Kotlin · Zig2022
  2. 14mesh-dnsTransparent sharding proxy for Postgres with online resharding.C++ · Rust2021
  3. 15pixelsortAn HTTP cache for edge workers with stale-while-revalidate semantics.Elixir · Go2020
  4. 16kvstoreDistributed tracing toolkit that keeps overhead under one percent.Zig · TypeScript2019
  5. 17rate-gateRuns untrusted WebAssembly with strict CPU and memory budgets.Rust · Python2018
  6. 18jsonpath-rsA Markdown editor that stays fast on 50k-line documents.Go · Kotlin2017
  7. 19deploybotA minimal Raft implementation with a deterministic simulation test harness.TypeScript · C++2025
  8. 20flamegraph-webTransparent sharding proxy for Postgres with online resharding.Python · Elixir2024
  9. 21bloomkitAn HTTP cache for edge workers with stale-while-revalidate semantics.Kotlin · Zig2023
  10. 22geo-tilesDistributed tracing toolkit that keeps overhead under one percent.C++ · Rust2022
  11. 23ci-lensRuns untrusted WebAssembly with strict CPU and memory budgets.Elixir · Go2021
  12. 24s3-syncA Markdown editor that stays fast on 50k-line documents.Zig · TypeScript2020
03

Experience

  1. 2021 — Now

    Staff Engineer, Database Platform

    Northwind DataAmsterdam

    Owned a critical piece of the platform end to end, from design docs and capacity planning to on-call and migrations. Worked closely with product teams to make the infrastructure invisible.

    • Led a zero-downtime migration of 40 TB across regions
    • Cut p99 query latency from 180 ms to 35 ms
    • Mentored five engineers into senior roles

    Postgres / Go

  2. 2019 — 2021

    Senior Engineer, Storage

    Northwind DataAmsterdam

    Owned a critical piece of the platform end to end, from design docs and capacity planning to on-call and migrations. Worked closely with product teams to make the infrastructure invisible.

    • Led a zero-downtime migration of 40 TB across regions
    • Cut p99 query latency from 180 ms to 35 ms

    Postgres / Go / Rust

  3. 2017 — 2019

    Senior Software Engineer

    Kite PaymentsLisbon

    Owned a critical piece of the platform end to end, from design docs and capacity planning to on-call and migrations. Worked closely with product teams to make the infrastructure invisible.

    • Led a zero-downtime migration of 40 TB across regions

    Postgres / Go / Rust / Kubernetes

  4. 2015 — 2017

    Software Engineer, Infrastructure

    Kite PaymentsLisbon

    Owned a critical piece of the platform end to end, from design docs and capacity planning to on-call and migrations. Worked closely with product teams to make the infrastructure invisible.

    • Led a zero-downtime migration of 40 TB across regions
    • Cut p99 query latency from 180 ms to 35 ms
    • Mentored five engineers into senior roles

    Postgres / Go

  5. 2014 — 2015

    Backend Engineer

    Loop LogisticsLisbon

    Owned a critical piece of the platform end to end, from design docs and capacity planning to on-call and migrations. Worked closely with product teams to make the infrastructure invisible.

    • Led a zero-downtime migration of 40 TB across regions
    • Cut p99 query latency from 180 ms to 35 ms

    Postgres / Go / Rust

  6. 2013 — 2014

    Software Engineer

    Brightline GamesLisbon

    Owned a critical piece of the platform end to end, from design docs and capacity planning to on-call and migrations. Worked closely with product teams to make the infrastructure invisible.

    • Led a zero-downtime migration of 40 TB across regions

    Postgres / Go / Rust / Kubernetes

  7. 2012 — 2013

    Engineering Intern

    Brightline GamesLisbon

    Owned a critical piece of the platform end to end, from design docs and capacity planning to on-call and migrations. Worked closely with product teams to make the infrastructure invisible.

    • Led a zero-downtime migration of 40 TB across regions
    • Cut p99 query latency from 180 ms to 35 ms
    • Mentored five engineers into senior roles

    Postgres / Go

  8. 2011 — 2012

    Research Assistant

    University Systems LabLisbon

    Owned a critical piece of the platform end to end, from design docs and capacity planning to on-call and migrations. Worked closely with product teams to make the infrastructure invisible.

    • Led a zero-downtime migration of 40 TB across regions
    • Cut p99 query latency from 180 ms to 35 ms

    Postgres / Go / Rust

04

Capabilities

Languages
RustGoTypeScriptPythonKotlinC++SQLElixirZigBash
Databases
PostgresSQLiteRedisClickHouseFoundationDBKafka
Infrastructure
KubernetesTerraformNixAWSGCPEnvoyeBPF
Practices
Design docsIncident reviewCapacity planningMentoringProperty testing
Observability
OpenTelemetryPrometheusGrafanaJaegerpprof
05

Education

  1. 2010 — 2012

    MSc, Computer Science, Distributed Systems

    Delft University of Technology

    Thesis on clock synchronisation in geo-replicated databases.

  2. 2007 — 2010

    BSc, Computer Engineering

    Instituto Superior Técnico

06

Writing(12)

  1. Dec 2025What happens to your p99 when one disk gets slowInfoQ
  2. Nov 2025Retries are a distributed systems problem
  3. Oct 2025A field guide to clock skew
  4. Sep 2025Writing design docs people actually read
  5. Aug 2024Deterministic simulation testing, in practice
  6. Jul 2024Postgres replication slots and how they biteInfoQ
  7. Jun 2024What happens to your p99 when one disk gets slow
  8. May 2024Retries are a distributed systems problem
  9. Apr 2023A field guide to clock skew
  10. Mar 2023Writing design docs people actually read
  11. Feb 2023Deterministic simulation testing, in practiceInfoQ
  12. Jan 2023Postgres replication slots and how they bite

Not looking right now

Let’s talk.