AleksandrKadesnikov

Neurowerk Agent

Own product, in production

An AI agent platform for direct messages. Each client gets an agent that answers on Instagram and Telegram, books appointments, captures leads and calls a person when it should.

1. Instagram, Telegram
2. Agent
3. Calendar, CRM, operator

• Books straight into Google Calendar, with no double bookings
• Hands a conversation to a person with a written summary
• Admin panel with versioned prompts, a test chat and a monthly budget per client
• Unit tests for the platform, scenario evals for each agent on the live model

Python, FastAPI, Postgres, Redis, Docker, MCP. Runs on Kimi K2.6 or Claude.

Call follow-ups

Insurance agency, Germany

A brief arrives before each client call. After the call, action points are drafted from the transcript using prepared templates.

1. Calendar
2. Brief
3. Transcript
4. Action points

n8n, Telegram, Google Calendar, Gmail, Keep, Drive.

Instagram assistants

Dance studio, massage therapist

Assistants that answer client questions in Instagram direct messages, so the owners can stay with their clients.

1. Instagram
2. ManyChat
3. Voiceflow
4. Reply

ManyChat, Voiceflow.

Control automation

Global insurer

A group-wide control framework where checks run automatically through system integrations and GenAI. I was lead architect and deputy product owner.

1. Source systems
2. ServiceNow
3. GenAI checks
4. Control results

ServiceNow, system integrations, GenAI.

Market signals

Experiment

A Telegram bot that streams live Polymarket trades, profiles each wallet and alerts on unusual bets. It tracks the hit rate of every signal to tune itself.

1. Trade stream
2. Wallet profile
3. Score
4. Telegram

Python, WebSocket, SQLite.

Code on GitHub