PABLO.ENGINEERING
GITHUB
SYSTEMS & LAB EXPERIMENTS: ACTIVE
// WARSAW, PL ·

AUTONOMOUS AGENT
SYSTEMS ARCHITECT
& AI-VIBECODER

I design multi-agent systems, event-driven architectures and async job queues. Solid software engineering (Python, Node.js/TypeScript, SQLite WAL) paired with rapid LLM-assisted prototyping.

EXPLORE PROJECTSREAD THE DEVLOG
TELEMETRY // ORCHESTRATOR
v2.4 // SNAPSHOT
CORE RUNTIMENode 22 LTS / Python 3.12
ACTIVE AGENTS8 Specialized Roles
STORAGE & QUEUESQLite (WAL) + Event-Driven
DIGEST ENGINEAutomated Git Pipeline
Daily Auto-Digest: Git commits are synthesized into devlog articles every 24 hours.
// 01. LAB PROJECTS

FEATURED SYSTEMS

[ 04 CASE STUDIES ]
CORE SYSTEM

AGENT OPERATIONS CENTER (AOC)

Control panel steering a team of 8 specialized AI agents (PM, devs, QA, AppSec). Instead of slow polling, tasks spawn in the background in 50ms (Node.js child.unref()) with SQLite WAL ensuring lock-free writes. Outcome: the UI never lags and the whole system runs smoothly on a single budget VPS without costly Redis clusters.

TEST SUITE
253 TESTS PASS
ARCHITECTURE
EVENT-DRIVEN
SPECIALIZATIONS
8 ROLES
#Next.js 15#TypeScript#Anti-Slop Guardrails#SQLite WAL#Event Broker#SSE Streaming#Docker

APPSEC & QUALITY GATE

A strict engineering guardrail for code written by AI agents. Bans compiler bluffing (type loopholes in TypeScript), enforces runtime Zod validation, and runs an independent critic on every commit. Practical value: agent code performs reliably like Senior dev output, catching bugs and vulnerabilities before they ever hit production.

[AUTO-AUDIT] DAILY VIA CI
  • Anti-Slop TS Guardrails
  • SAST AST Inspection
  • Strict XSS Sanitization
// AI AGENTS // GARMIN .TCX TELEMETRY

AI COACHES (TRAINANBOT & PABLOTRAINER)

TELEGRAM LIVE

Two Telegram coaching bots tailored to real physiology instead of generic online plans: TrainAnBot adapts to female pelvic biomechanics and PubMed evidence, while PabloTrainer manages fatigue across boxing, gym, and running. They parse raw Garmin .tcx XML files in 2 seconds, calculating heart rate zones to prevent injuries without manual data entry.

#AI Agents#OpenClaw#Telegram Bot#Garmin .TCX#PubMed API#Combat Conditioning
// AI AGENT // CODED PEDAGOGY

PROFECLAW - SPANISH TUTOR

A digital Spanish tutor on Telegram built on structured pedagogy: corrections are capped at 2 per turn so learners never feel overwhelmed, an error counter automates spaced repetition, and idioms are translated sense-for-sense. An automated eval suite guarantees the bot never hallucinates false word etymologies.

#AI Agent#OpenClaw#Pedagogy Engine#Spaced Repetition#Eval Suite
// R&D LAB

DEEP CS & VIBECODING LAB

My daily playground for Python and TypeScript exploration: data structures, Linux server hardening, and custom developer tooling. Daily Git commits are automatically synthesized by a dual-Gemini pipeline into rigorous devlog posts, fact-checked against source commits before human review.

#Algorithms#VibeCoding#Node.js 22#Daily Standup
// 03. ARCHITECTURE & CAPABILITIES

ENGINEERING TIMELINE

[ TIMELINE & STACK ]

Built a control center to steer a swarm of 8 specialized AI agents (PM, coders, QA, AppSec). Replaced sluggish database polling with instant detached background execution (Node.js child.unref()) backed by SQLite WAL. Practical outcome: task latency plummeted from 4s to 50ms, the UI stays lightning-fast, and the entire system runs reliably on a single lightweight VPS without heavy Redis clusters.

Next.jsTypeScriptSQLite (WAL)Event-DrivenDockerSubagents
// 04. LAB / CONTACT

HAVE A TRICKY
AGENT PROBLEM?

Show me a problem for AI agents to solve - event-driven architecture, data analysis, or unreliable automation. I like hard cases. I reply to every sensible message.

contact@paterski.com
GITHUB PROFILE