Senior citizens comprise a fifth of our population, and are among the most vulnerable to financial scams. Tragically, victims are often left defenceless by the rapid pace of technological advancement, and cultural expectations. The average elderly victim suffers $33,000 in damages and lives shorter due to the stress.
Guardo is here to protect your family.
Data Collection (BrightData → Perplexity) • BrightData triggers multiple Perplexity searches daily for scam-related content • Results are aggregated into a JSON document database (scam_news_YYYYMMDD.json)
Pattern Extraction (ZeroEntropy) • ZeroEntropy AI analyzes the collected JSON data • Extracts elderly-specific patterns
Knowledge Indexing (LlamaIndex) • Takes ZeroEntropy's extracted patterns • Builds a searchable vector database of scam patterns • Updates daily with new scam trends
Real-time Protection (VAPI) • When a call comes in, VAPI triggers • Queries LlamaIndex to match call content against known scam patterns • Makes real-time decisions: • Block high-risk scam calls • Transfer to family for suspicious but uncertain calls • Allow legitimate calls through
Post-call Analysis (Senso AI) • Analyzes completed call transcripts • Generates family alerts with detailed summaries • Creates audit trail for pattern improvement
Frontend • Guardo.ca acts as the dashboard, monitoring it all.
Challenges we ran into
Vapi was the cornerstone of our product, but dealing with phone calls in production is hard. We really wanted to implement warm transfers to family members, but the support for conference calls and third party warm transfers was quite limited.
Some of the agents we wanted to use were difficult to implement. The demo does not include some planned agent integrations like Senso and Llamaindex.
Accomplishments that we're proud of
This was our first time integrating a multi agent loop. It was fun to see them all working together!
What's next
This was a problem we had in our mind for years, and it will only get worse as time goes on. As the technology improves, especially with providers like Vapi, we want to implement our full vision of proactive defence. We even thought about going on the offensive and hunting down telephone scammers.
Built With
- brightdata
- python
- replit
- senso
- vapi
- zeroentropy
Log in or sign up for Devpost to join the conversation.