Graphs beat forms.
Any domain where relationships drive revenue, from sales to recruiting to partnerships, rewards graph thinking over record thinking. The data model is the strategy.
The Lab · Case Study 04
Your network is a market you already own.
Everyone in business says relationships are everything, then updates their CRM twice a quarter. Meanwhile the network decays silently: people change jobs, signals go unseen, warm paths go cold, and the introduction you could have made next month becomes impossible next quarter.
Relmint is relationship intelligence for people who sell. It watches the signals across your professional network and recommends the next move. You send. It never does.
LiveEarly product
The business case: CRM decays because it asks too much
The CRM was supposed to be the system of record for relationships. It became a system of homework: fields to fill, updates to log, hygiene nobody does. The data rots because entry is a tax on the person who benefits least from paying it.
The fix is not a nicer form. It's inversion: let the system do the watching. Public signals like job changes, company news, and posts are already out there, changing daily. A relationship platform should ingest that stream, map it onto your actual network, and tell you what changed and why it matters.
That's Relmint: an intelligence graph over your professional relationships, built and maintained by the system, surfaced as recommendations at the moment they're actionable.
The warmest introduction you will ever get is the one the data spotted before you needed it.
What it does
Signal watching: public signals across your network, from role changes to company events to posts, detected and attached to the right people, with sources.
Warm paths: the graph infers introduction routes. Who you know who knows who you need. The shortest path to a decision-maker is a mutual friend, not a cold email.
Clean import: contacts arrive messy from everywhere. Fuzzy matching and duplicate resolution fold them into one coherent graph instead of five versions of the same person.
Recommended next moves: outreach suggestions grounded in relationship history, drafted for review, sent by you. The same proposals-only doctrine as every system in this lab.
Under the hood
- Platform
- A FastAPI backend with a React frontend, PostgreSQL in production
- Intelligence
- Neural search over public sources plus LLM synthesis to extract, deduplicate, and rank relationship signals
- Graph
- Relationship graph with explicit and inferred edges. The data model that makes warm-path inference possible
- Safety posture
- Viewer and recommender by design: it watches public signals and drafts recommendations. Every outbound message is human-sent
What this proves for client programs
Signal-driven GTM is here.
Timing, not volume, is the modern outreach lever. Systems that surface the moment beat sequences that spray the month.
Trust-safe automation sells.
Same doctrine as the AI SDR pattern: watch, recommend, hand off. Enterprises adopt faster when the AI visibly cannot embarrass them.
The bridge to your program
Selling into networks you can't see?
Most pipeline problems are visibility problems wearing a quota costume. I help teams instrument their market with signals, graphs, and the governance to act on them safely.