AI Developer · Novosibirsk, Russia · Building autonomous AI systems

I design and build production-grade AI agent systems — from multi-bot Telegram workflows to autonomous task automation pipelines running on Claude, Notion, and local Mac hardware.

• 🤖 Multi-agent systems — OpenClaw-based bots with clear role separation (secretary + executor model)
• 🧠 Autonomous pipelines — Self-running workflows: Notion ↔ Calendar sync, task routing, sub-agent spawning
• 📬 Telegram AI Bots — Meeting processors, task managers, voice-to-task automation
• ⚙️ Local-first stack — Mac Mini M4 + local LLMs (Ollama) + cloud fallback (Claude Sonnet/Haiku)
• 🔧 Open Source — Contributing to OpenClaw ecosystem, publishing tools and agent configs

My personal AI operating system running 24/7 on Mac Mini M4:

User (Telegram)
    │
    ▼
┌─────────────┐        ┌──────────────────────────────────┐
│  Assistant  │───────▶│  Notion (Tasks DB + Calendar DB)  │
│  (Secretary)│◀───────│  Source of truth for all tasks    │
└─────────────┘        └──────────────────────────────────┘
    │                               ▲
    ▼ sessions_send                 │ auto-sync every 15 min
┌─────────────┐        ┌──────────────────────────────────┐
│     CEO     │        │  notion_sync.py + macOS launchd   │
│  (Executor) │        │  → Apple Calendar (.ics)          │
│  sub-agents │        └──────────────────────────────────┘
└─────────────┘
    │
    ├── research_agent   (web search, analysis)
    ├── content_agent    (writing, copywriting)
    ├── code_agent       (scripts, automation)
    ├── analysis_agent   (data, decisions)
    ├── notion_agent     (Notion read/write)
    └── strategy_agent  (planning, roadmaps)

• Receives all incoming requests via Telegram
• Clarifies: what, deadline, priority (P1/P2/P3), context
• Creates entries in Notion Tasks DB + Calendar
• Notifies CEO via sessions_send
• Confirms completion back to Sasha
• Does NOT execute tasks — intake only

• Reads task from Notion on notification
• Spawns specialized sub-agents via sessions_spawn
• Sends result directly to Sasha in Telegram
• Reports completion → Assistant marks Done in Notion

Full agent configs: openclaw/agents/

SaaS Telegram bot for meeting processing. Transcribes audio/video (Whisper), extracts summaries and actionable tasks via LLM, subscription tiers with SQLite.

OpenClaw plugin toolkit — reusable skills and agent components.

Curated collection of OpenClaw skills and use cases.

botus-ai/
├── README.md                         ← You are here
├── Scripts/
│   ├── notion_sync.py                ← Notion → Calendar auto-sync (15 min)
│   ├── install_sync.sh               ← One-command setup
│   └── com.botus.notion-sync.plist   ← macOS launchd config
├── openclaw/
│   ├── agents/
│   │   ├── assistant/SOUL.md         ← Secretary agent config
│   │   └── ceo/SOUL.md               ← Executor agent config
│   └── ARCHITECTURE.md               ← Full system architecture
└── docs/
    └── SETUP.md                      ← Setup guide

# 1. Clone repo
git clone https://github.com/botus-ai/botus-ai.git
cd botus-ai

# 2. Set your Notion token in Scripts/notion_sync.py
# Create integration at: https://www.notion.so/my-integrations

# 3. Connect integration to your Notion databases
# (Tasks DB + Calendar DB → ••• → Connections → Add connection)

# 4. Install and start
bash Scripts/install_sync.sh

This registers a macOS launchd agent that syncs every 15 minutes and generates an .ics file for Apple Calendar subscription.

• Python 3.9+ — macOS compatible, no venv required
• Telegram Bot API — python-telegram-bot
• LLMs — Claude Sonnet/Haiku (Anthropic), Ollama (local)
• Databases — Notion API, SQLite
• Automation — OpenClaw framework, macOS launchd
• Calendar — iCalendar (.ics), Apple Calendar
• Hardware — Mac Mini M4

• Email: sasha0046657@gmail.com
• Location: Novosibirsk, Russia 🇷🇺
• Open to: Consulting, freelance, co-founding

Building AI that actually works in production 🚀