← back to the storm

About

I quit my job a while ago. The plan was a break — some freedom, some rest. Instead, the nerding kicked into a higher gear than ever, and it's been far more fun than any holiday.

Then came drylane.io and The Dry Stack: a database, a message layer, and a serializer — fused into one stack, written from scratch in C++ with zero-copy C# wrappers, plus a real-time audio engine for good measure. The core came together in about a month.

The benchmarks so far? Very satisfying. Not impressed? Let's run them on your hardware. 😃

40+ years building
3 core engines, fused
~1 month to a working stack
millions/sec ingest
microseconds per query
How it's built

Every architecture and design decision is mine. AI comes in as an iterative accelerator, under tight technical direction — a tool that lets one experienced engineer move at the speed of a team, not a substitute for knowing what you're doing. That's how a stack this broad shipped this fast. And the ML fundamentals are old ground here — a from-scratch neural-network engine in C++, with backpropagation and live learning-rate visualization, dates back to the DTU years.

Forty years in

None of this came out of nowhere. The first program came at six, the first game at nine, and not much has stopped since — from the Amiga demoscene and 68k assembler, through real-time audio DSP at DTU, two decades of embedded payment and industrial systems, and a from-scratch time-series database now being rolled out as the standard for vessel automation and event data at a leading player in the marine industry. The Dry Stack is what all of that compounds into. (Unda, the audio engine, even closes the loop back to those university DSP projects.)

What I'm building
Facts
Time-series / event store answering in microseconds — SIMD query surfaces over billions of records, append-only and tag-oriented.
Where you'd reach for InfluxDB or Timescale — on one box, far faster.
Nio
Network I/O built around bidirectional events, messages, batching and streaming — tens of millions of messages/second.
Where you'd reach for Kafka — without the copies and the hops.
Zao
Generator-based serializers at memcpy speed with minimal memory overhead. Zeroes And Ones.
Where you'd reach for Protobuf — at the speed of a raw memory copy.
Unda
A real-time, low-latency audio engine — DSP building blocks for voice and guitar-effect processing.
It already powers a peer-to-peer VoIP proof-of-concept built on Nio. Once you have transport, audio, and a curiosity about latency, it has to be done.
Where's it going? Not entirely sure. But the direction is clear: fast, fun, cool, and good for the world.

Also open to remote engagements — the challenging, genuinely interesting, fairly compensated kind. If you've got a hard problem in performance, systems, low-level, or real-time, that's exactly my kind of problem.
info@drylane.io · LinkedIn · X