Reach me:
Finished, working tools with setup guides, test suites and containerized deployments:
- voice-agent
Real-time outbound voice agent: live speech-to-speech lead qualification via OpenAI Realtime protocol, state-machine dialogue flow, SIP telephony & signed webhooks.Python·FastAPI·OpenAI Realtime·SIP·WebSockets·Docker - caseflow-ai
Human-in-the-loop document ingestion pipeline: automated PDF splitting, layout OCR, schema extraction, and deterministic risk routing with a web review UI.Python·Tesseract OCR·PyMuPDF·Docker - presales-planner
Telegram bot that auto-schedules a team's tasks across people and days — respecting capacity, dependencies and deadlines — with a FastAPI web admin. Deterministic greedy solver over a critical-path graph; Claude intent parsing; text + voice task capture.Python·aiogram·FastAPI·PostgreSQL·SQLAlchemy·NetworkX·Claude - ai-redaktor
Telegram bot that proofreads and edits Russian text — spelling, punctuation, grammar and unjustified anglicisms — in a single structured Claude call, returning a corrected file, a summary, and a full list of fixes. One-command Docker run; tokens via.env.Python·python-telegram-bot·Anthropic Claude·pandoc·Docker - notion-telegram-notifier
Notion → Telegram push notifications for teams on Notion guest seats, where native notifications stop working. Polls Notion, diffs snapshots, and DMs the right person on assignment or status change.docker compose up -d.Python·aiogram·Notion API·SQLite·Docker - SpotiDL DJ
CLI that downloads Spotify playlists to MP3 with correct-track matching — you get the exact song and version you named, never a random remix or live cut — with full metadata and album art. Step-by-step setup incl. Spotify API keys.Python·yt-dlp·FFmpeg·Spotify API
I spend real time mapping the dev & AI-tooling landscape and sorting what's worth keeping into curated lists I actually use:
The AI tools I work with every day to ship real software:
