Engineering researcher. APIs, databases, infrastructure, systems design.
The schoolroom story of telecommunications begins with Samuel Morse and the 1844 Washington-to-Baltimore line carrying "What hath God wrought?" The story usually skips fifty years of
Most API teams configure Cache-Control once when they ship the API and never touch HTTP caching again. The result is one of two failure modes: caches that lie
Most product teams pass through a phase where they build an internal SDK for their own API. The intentions are good: to encapsulate authentication, to provide typed bindings,
The schoolroom story of multi-region SaaS goes through a globally distributed database with strong consistency, automatic failover, and transparent routing. The cost of that story in 2026 is
Probabilistic data structures trade exact answers for orders-of-magnitude memory and speed wins, and they are the right tool more often than developers expect. Bloom filters answer set-membership in constant time with controlled false-positive rates, and HyperLogLog estimates cardinalit...
Backups and archives are different products that are often built with the same infrastructure, and the conflation produces both expensive backups and unrecoverable archives. The honest distinction is operational vs historical, with different retention policies, different access patterns...
Notification systems start as a single email send and end as multi-channel infrastructure with preferences, throttling, and delivery tracking. The patterns that survive the growth path treat notifications as events and channels as adapters, with user preferences as a first-class data mo...
API quotas are the load-bearing constraint of any SaaS pricing model. Get them wrong and either the free tier abuses your infrastructure or the paid tier feels arbitrarily limited. The honest framework separates rate limits from quotas from billing meters, and treats the metering layer ...
Most teams reach for Elasticsearch the moment search appears on the roadmap, even though three earlier options would have shipped faster and stayed simpler. The escalation order matters because the wrong starting point bakes in operational complexity that compounds.
Partitioning is one of those Postgres features that gets reached for whenever a table is large, even though large is not the same as needs-partitioning. The honest framework is workload-shaped: certain access patterns benefit dramatically, others get worse, and the operational cost is permanent.
The Cloaca Maxima drained the Roman Forum from 600 BCE and is still in service. Bazalgette's London sewer of 1865 ended cholera in the city. Edwin Chadwick spent decades fighting public health authorities who insisted that disease came from bad smells. The history of sewer engineering i...
The architecture conferences talk about horizontal scaling. The reality of running a small SaaS is that vertical scaling solves most problems for years longer than anyone expects, and the question of when to switch is mostly about predicting future growth that is hard to predict.