WEBI

Walk onto any asset-management or family-office desk in Paris, Geneva, or Luxembourg and you will find a structured-products analyst three monitors deep, scrolling hundreds of Bloomberg messages and bank emails. Each one pitches a structured bond product that has to be cross-checked against the fund’s mandate, risk tolerance, sector exposures, and a dozen mostly-implicit constraints. Talking to people who do this job during the hackathon, the pattern was stark: roughly 70% of their day goes to manually filtering offers that will never match, and real opportunities get buried in the wrong inbox.

WEBI (“We buy”) is the SaaS platform that filters deal flow and helps shape the right products, an AI copilot that turns fragmented bank deal flow into a curated, mandate-matched, negotiation-ready pipeline. It works in three layers:

• Intelligent filter: it ingests emails, Bloomberg chats, and term-sheet PDFs, uses Gemini 2.0 Flash to extract a 17-field structured representation of each product (issuer, underlying, coupon, barrier, maturity, ratings, currency, ESG flags, and more), and a FAISS vector index to match it against each fund’s codified mandate. Hard violations are auto-rejected, matches above 90% reach the analyst with a transparent score and explanation, and near-misses are kept for negotiation.
• Negotiation engine: about 80% of structured products are adjusted after discussion with the issuing bank, so WEBI turns a near-miss (say a 4% coupon against a 4.5% mandate) into a drafted counter-offer that tells the analyst exactly which parameter to push, by how much, and why the fund is justified in asking.
• Market discovery: a secure, anonymous cross-fund signal layer. If one family office receives a deal that better fits another’s mandate, WEBI routes the connection, and it runs every incoming deal against an entire mandate database at once, built for the multi-family offices where a single analyst juggles dozens of mandates.

Built with Gemini 2.0 Flash, Cerebras, FAISS, Python, and TypeScript.

Built in 24 hours at the Paris Fintech Hackathon (Solve with AI) at HEC Paris, France’s premier student Generative AI and Fintech hackathon: 150 selected hackers, hosted by SP AI (Sciences Po), ASES France (HEC & Stanford), and Google Developer Groups. We pitched WEBI to a jury of VCs and founders. No prize this round, but a crash course in finance, negotiation, and shipping a product overnight.

Team: Louis Berthier, Emile Jouannet, Mathilde Anita Thierry, Elie de Baudus, and Roland Sosin.

Pitch