Inspiration

While half the team are in Taiwan to do validation with manufacturing, a lot of the engineers in hardware companies we met wanted to get better landed rates or better suppliers in order to make more attractive prices to end consumer. So in the first few days ,we validated with Taiwan top system integrators and corporates such as Systex and Hotai to understand their supply chain and came up with advanced sourcing.

What it does

Advanced sourcing not only includes agents that check your manufacturing files for any missing details, it also checks the prospective suppliers based on given requirements (financial standings, expertise, years of operations, inventory list and part documentation, MOQ and landed cost, sample cost, previous clients etc.) by first looking through publicly available sources and deploying agents to message on their behalf

It also can use those same agents to ask for quotes and negotiate better prices with valid justification to deliver the final quotation within 1 week.

How we built it

We first use Exa to manage the scraping capability of publicly available sources. Then we implemented a intelligent multi-engine search skill built specifically for agents. It automatically routes queries through Bing, DuckDuckGo, and Tavily, and can bypass VPN restrictions to reach chinese sites. We then use the Chatgpt to reason and also come up with the email before using Resend API to connect to our email paired with a framework and benchmark for evaluating LLM/VLM agents in realistic buyer–seller commerce. It extends negotiation beyond text-only, bilateral price haggling into multimodal, multi-dimensional contract negotiation.

For selection and comparison, we use another agent for document OCR, vendor scoring, and web research. Allow both user and the agent to analyze bids, extract structured data, and make decisions.

Challenges we ran into

Definitely it would have to be coordinating all the different agents to work together, while there are other public github repos for us to look into and gain inspiration, we still have difficulty to branch them together and work with our current ChatGPT API a the main reasoning thing holding all else together.

Accomplishments that we're proud of

3 POC meeting with Taiwanese Enterprises (Systex, Hotai, EliteGroup Computer Systems CO. Ltd.) Direct $30,000 Angel SAFE Investment Offer

What's next for Advanced Sourcing Agent

Value add with our current buildables lineup and prepare for a b2c launch for non-technical to vibe-engineer the safe and proper way.

Built With

  • chatgpt
  • cursor
  • resend
  • supabase
  • vercel
Share this project:

Updates