Lead → Outreach
Strategy & Knowledge • Work sample
Live: outreach.bastian-brand.com
What it is
The research-and-first-touch loop an SDR does by hand, automated. Give it a target company and it reads their website and recent news, scores how well an offering fits their likely needs, pulls out the concrete signals worth hooking the outreach on, and drafts a personalized email and LinkedIn note — grounded in what it actually found, not a mail-merge. The offering it pitches defaults to my own consulting, so the demo doubles as its own first touch: it drafts the message a prospect would receive.
Sample companies are pre-run and cached; “run your own” is live.
How I built it
Firecrawl resolves a company name or domain to its site, scrapes it, and pulls a small set of recent-news results. An LLM then produces a structured package in one pass — a snapshot, a 0–100 fit score with a one-line rationale, three to four specific signals lifted from the company’s own material, and a short email + LinkedIn note that reference those signals. The prompt is strict about grounding: it may only cite real specifics (a product, a market, a recent acquisition) and must stay honest when the material is thin — the difference between personalization that lands and spray-and-pray that gets deleted.
Highlights
- Real personalization: every draft references specific, verifiable signals
- Fit score + rationale, so a rep knows whether to send before what to send
- Grounded by design — the model is told not to invent facts about the company
Stack
Python · Streamlit · Firecrawl · OpenRouter (LLM)