← All work
TGLeads
87–92%filtering accuracy
12–20warm leads per week
+28%revenue growth in HR vertical
The challenge
- The client manually watched 40 chats for relevant project requests.
- This split attention, wasted time on irrelevant messages, and meant multi-hour reaction times — losing potential clients.
Our solution
- Built AI-based message analysis across target Telegram chats, with typical reaction delay around 15 seconds (up to 2 minutes).
- Scaled to monitor up to 1,000 channels at once, with message queues and dedicated processing/filtering infrastructure that adapts to new communication patterns.
- Added an analytics and operations dashboard, plus parsing for both open and closed channels/chats.
Engineering challenges
- Keeping AI classification accurate as chat topics and slang shift.
- Scaling real-time monitoring to 1,000 channels without losing responsiveness.
Inside the product
A look at what we shipped.
Drop a screenshot here
Drop a screenshot here
The outcome
A happy customer.
The client is now alerted within a minute of a lead posting in any of 50 monitored chats, and assistant productivity rose 30%.
87–92%filtering accuracy
12–20warm leads per week
+28%revenue growth in HR vertical
More from Generative AI
All work →
Generative AI
Smmik.online
AI editor that curates, rewrites and auto-publishes a Telegram news digest on schedule.
100concurrent users load-tested
Generative AI
Face Recognition Service
A sub-second face-similarity search engine built to power DIVO’s paid Face Search feature.
<1ssearch across millions of faces