Peleke Sengstacke's Work | ContraWork by Peleke Sengstacke
Peleke Sengstacke

Peleke Sengstacke

AI Systems Architect: Agentic AI, Security, and Learning

Ready for work

Peleke is ready for their next project!

LinWheel in action: drafting, reshaping, and scheduling a LinkedIn post from Claude Code...With carousels! No context switching; full control over what goes out (and when) via human-in-the-loop gates; and automated asset creation. There's lots more cooking behind this one—ping if you're interested in joining the beta group; pots are limited!
1
2
42
Cover image for If your agents restart from
If your agents restart from zero every session, can't remember what they learned yesterday, or you have no idea why they break...That's what I fix. I do: - Security audits for agent deployments - Set up and deploy AI agents across your messaging platforms (Slack, Discord, Telegram, WhatsApp) - Custom workflow automation connecting your existing tools
1
62
Cover image for AI language tutor that actually
AI language tutor that actually understands linguistics. Classifies learner errors by real linguistic taxonomy (not just "wrong"), continuously calibrates proficiency, and adapts retrieval spacing based on what you're actually struggling with.          Currently focused on Old Norse. Built with qortex, Next.js, Vercel, Google Cloud Platform.
1
32
Cover image for MCP server that turns your
MCP server that turns your engineering work into a feedback loop. Ambient journal capture on every commit, multi-persona code review (the "gauntlet"), skill extraction from patterns across entries, and reward signals that close the loop between what you shipped and what you learned. Runs in every project via MCP: zero friction, always on. 
0
26
Cover image for Personal agent runtime with per-tool
Personal agent runtime with per-tool sandboxing. Runs across 18+ messaging platforms (WhatsApp, Telegram, Slack, Discord, Teams, etc.) with voice interaction, scheduled automations, and custom tool orchestration. Every tool executes in its own sandbox; the agent can't escalate beyond what you grant it. Built on OpenClaw.
0
32
Cover image for Multi-agent LinkedIn content pipeline.

Extracts a
Multi-agent LinkedIn content pipeline. Extracts a voice profile from your existing writing, generates posts across six rhetorical angles, renders branded carousels, and handles scheduling + analytics. 17 MCP tools, 933 passing tests, runs as an extension to any MCP client in existence. Built with TypeScript, Next.js, Claude API.
1
37