The systems we've architected and operate: proof that our advice comes from running production systems, AI included, not writing decks about them. Five proofs, equal weight, no favourites.
Five panes of evidence: the systems our practice is built on and the work we deliver. Drag or swipe the gallery to explore the systems we run. The full detail for each follows below.
What each system does, the buyer fear it retires, the stack behind it, and the highlights that matter.
An autonomous AI engineering agent we built on Claude and run in production.
When we advise on AI, this is what's behind it. Give Grace a goal and she breaks it into phases, hands each to an AI coding agent, supervises them in parallel, verifies the work, and reports back, all controllable from our phone. Cost ceilings and human sign-off on anything destructive are built in, and unattended mode is opt-in rather than the default. We live with the governance trade-offs of production AI daily, so the guardrails we design for you are ones we've already had to get right ourselves.
"AI is a slick demo, not a colleague." Grace is proof that an agent can hold real responsibility, and that we know how to build, govern, and trust one.
The AI-operated ops & intelligence platform we run our entire practice on: the same class of system we stand up for clients.
It runs the business end to end (idea capture, sales pipeline, build board, proposals, contracts, engagement and budget tracking, and revenue), alongside an AI learning center, intelligence feeds, and a knowledge base that refresh on their own daily. Agent-operated: Grace reads and writes through a governed, key-authenticated API that keeps structure separate from live data, the human-plus-agent operating model we help clients build, running in our own production.
"Advisors who don't use what they sell." We live inside our own platform daily, so our recommendations are operating experience, not theory.
Self-hosted language models for a privacy-sensitive environment, running at zero cloud cost. Data never leaves the building.
A self-contained AI environment where every model runs on local hardware, built for a setting where sensitive data and offline operation were non-negotiable. Users reach the tools through a browser with no accounts and no data leaving the premises. The architecture we built: Ollama serves local models on-premise while a Python backend handles routing, guardrails, and session management, the same pattern we apply anywhere data residency rules out the public cloud.
"Our data ends up training someone else's model." We design AI that earns its keep without ever surrendering custody of what makes you, you.
An automated pipeline we built to pull intelligence from many sources into one always-current feed.
It collects, summarizes, and scores relevant updates from RSS and web sources, then surfaces them in our internal tools hub, so we stay current on vendor announcements, industry news, and security advisories without checking dozens of feeds by hand. No servers to run: the whole thing runs on Google Workspace and serverless GCP components on a schedule, folding RSS, web scraping, AI summarization, and relevance scoring into a single view.
"Insight is manual, slow, and stale by the time it lands." We turn intelligence into an always-on utility rather than a fire-drill.
Domain-wide email export to standard formats for compliance and e-discovery.
A command-line tool we built for IT and compliance teams pulling email at scale from Google Workspace. Service-account delegation exports across an entire organization without touching user credentials, for legal discovery, regulatory holds, migration, or archival. Built for the audit trail: exports come in standard EML format with a CSV index for search, and timestamped runs with ZIP packaging support chain-of-custody requirements.
"Discovery and audits mean scrambling through inboxes by hand." We make bulk, defensible email export a single command, with the audit trail built in.
No decks. No proposals. A direct conversation with us about what you're building and whether we're the right partner.
Prefer email? contact@sgltech.io