based in Thessaloniki, Greece / GitHub since 2012 / 98 public repos

Spiros Raptis

Agentic Systems Engineer / AI Systems Builder / Past Trading-System Work

I build open-source tools, automation, and AI workflows that are inspectable, versioned, composable, and boring in the good way.

Small systems that remove real friction.

My projects usually start as a script that solves one sharp problem: queue visibility, faster screenshots into Codex, private media playback, historical markets snapshots, or a local dashboard that says what is actually true.

The pattern is simple: expose the state, keep the workflow inspectable, then earn the abstraction only when repeated use proves it belongs.

Practical tools with visible machinery.

coinbase-perps-lab

Past Rust lab and local dashboard for inspecting Coinbase INTX perpetual positions.

Repo

CodexScreenshotBridge

Swift menu-bar app that moves screenshots straight into a Codex workflow.

Repo

udacity-mentor-dashboard-extension

Browser tooling for mentor workflows, daily totals, and queue automation.

Repo

zt-report

Shell reporting for ZeroTier member status, aliases, IPs, and visibility.

Repo

crypto-finance

Past Python experiments around trading systems, research, and automation.

Repo

plex-open-web

Self-hosted Plex web UI for private browser playback.

Repo

Guardrails before fireworks.

  1. 01 Make the state visible.

    Dashboards, logs, receipts, and diffs beat vibes when systems get real.

  2. 02 Start with scripts.

    Automation should earn complexity by surviving repeated use.

  3. 03 Use AI without hiding the logic.

    Agents are useful when the boundary, evidence, and rollback path are obvious.

  4. 04 Treat markets work like software.

    Rules, snapshots, stops, and journals were the useful artifacts.

Python, JavaScript, Rust, Swift, Shell.

Linux Docker tmux CLI tools Browser extensions LLM workflows Trading systems Local dashboards

Send the repo, the constraint, and the thing that keeps breaking.

I am most useful around practical open source, automation, agent workflows, software with sharp enough edges to be worth improving, plus the occasional lesson from past markets tooling.