Data Slug

•

Vintage floppy disk on wooden surface — Data Slug blog header image

Journey over Data Tools and Exploration

About

Senior Director of Platform Engineering → AI builder → learning in public

TLDR: 5+ years running production infrastructure at scale. One year learning that building with AI is a completely different problem than building for it. This blog is how I close that gap in public.

The honest version

I’m a Senior Director of Platform Engineering at Cognitive Systems in Kitchener-Waterloo, Ontario. I lead an 8+ person team managing infrastructure for over 1 million deployed IoT devices. The stack runs from embedded device firmware through Linux, Python cloud services, and increasingly AI/ML tooling.

Platform engineering teaches you to build things that don’t fall down — reliability, observability, clean abstractions, no surprises at 3am. That discipline transfers well into AI infrastructure work. What it doesn’t prepare you for is the judgment layer: when should the model handle this? When should the human? What does good output actually look like? Those questions don’t have runbooks.

So I started building things to find out.

What I’ve shipped

  • Council — multi-agent boardroom where AI executives debate and generate policy. The legislative branch of an enterprise AI judgment system.
  • Policy Pal — policy Q&A bot demonstrating five enterprise AI patterns: PII scrubbing, grounded responses, guardrails, confidence scoring, audit logging.
  • Standup Buddy — async standup tool that detects when engineers are stuck through voice tone, not just words. Built in 10 hours, placed 8th at AI Agents Waterloo.
  • NanoClaw fork — containerized agentic AI with a voice layer added via skills architecture.

Dev Note: I shipped Standup Buddy at a hackathon on 10 hours of sleep debt. I broke my NanoClaw fork trying to add voice the fast way instead of the right way. The posts here are about what I actually learned — not a cleaned-up version of the process.

Why this blog exists

A place to explore data engineering tools and AI engineering — in public, with the failures included. I’m not an AI researcher. I’m not selling anything. I’m an engineering leader who manages production infrastructure by day and builds data and AI systems by night, and I find the space between those two roles genuinely interesting.

The audience I’m writing for: engineers and engineering leaders interested in how AI and data tooling actually works in practice — not the pitch deck version, not the cleaned-up tutorial version. If you’ve ever lost two hours to scope creep at a hackathon, or broken an upstream dependency by patching core instead of extending it, or wondered whether voice activity detection is actually a hard problem — you’re in the right place.

Posts ship when something is worth saying. Usually that means something went wrong first.

Where to find me

If you’re building something in the agentic AI, data engineering, or platform engineering space and want to compare notes — reach out. I’m genuinely interested in what people are finding in the wild.

William Suriaputra is a Senior Director of Platform Engineering with 5+ years of production infrastructure experience, currently exploring data engineering tools and AI engineering at the intersection of platform engineering and agentic systems.