Vendly

Most people have skills.Very few have clients. Vendly finds real businesses, researches them, ,generates the best pitch,writes an outreach email, and becomes your personal sales agent in seconds.

Comment

Inspiration - A few weeks ago, a friend and I were miserable spending most of our time manually scrolling Google Maps to find web development leads. We were blasting out generic cold emails that got basically zero responses, but personalizing every pitch took hours we simply didn't have. I realized we shouldn't have to choose between doing the work we love and doing soul-crushing administrative research just to find our next paycheck. That shared frustration inspired me to build Vendly AI: an autonomous, 24/7 sales team that not only finding leads, but does the heavy lifting of actually researching them and drafting perfect pitches.

What it does - Vendly AI discovers local or global businesses via an interactive map, and then uses ASI-1 to actively research their online presence and recent reviews. It then autonomously generates hyper-personalized cold outreach emails that perfectly align your specific skills with that business's real-time pain points.

How we built it - We built the backend with Python and FastAPI, and designed the frontend to handle server-sent events for a real-time streaming UI. We deeply integrated the ASI-1 API for both document parsing and active web-search inference to power the entire lead research engine.

Challenges - we ran into Handling real-time streaming data from the ASI-1 model into the frontend UI while maintaining strict JSON data structures was a complex challenge. We had to carefully prompt the system so it could stream its "thought process" smoothly before returning the final, enforced JSON payload containing the pitch and scores.

Accomplishments that we're proud of - We are incredibly proud of our zero-shot document parser, which correctly understands casual bios like "I build stuff out of wood" without relying on hardcoded keywords. Plus, watching the AI actively browse a local business's real website and point out their specific flaws in the generated email feels like absolute magic.

What we learned - We learned how powerful combining geospatial data with real-time AI web search can be for hyper-local intelligence gathering. We also realized how effective strict system prompting is at turning a language model into a reliable, deterministic backend service rather than just a chatbot.

What's next for Vendly - We plan to fully integrate the Google Maps API for even richer, global geographic business data and more accurate local discovery. We also want to add automated email follow-up sequences based on our open-tracking data, and support for connecting popular CRM platforms like HubSpot.

Built With

Share this project:

Updates

Submission history