# rallion / war room
GTM engine · Human-in-the-loop
Rallion
Live market signals in. Evidence-cited, human-approved outreach out.
- Signals → evidence-cited opportunities → human approval → end-to-end outreach
- Prospects qualified by an active-learning model: GP classifier with BALD acquisition
- Every prospect × channel is a thread with a guarded state machine — illegal jumps raise
Signals 3
Opportunities 1
Approved 1
Thread state
discovered
qualifying
qualified
contacted
engaged
suppressed · from any state
Approve campaign?
Data-platform modernisation
- 3 prospects · evidence cited
- Email first, LinkedIn optional
- Activity window 9–5 · per-channel caps
nothing sends without me
- Signals → evidence-cited opportunities → human approval → end-to-end outreach
- Prospects qualified by an active-learning model: GP classifier with BALD acquisition
- Every prospect × channel is a thread with a guarded state machine — illegal jumps raise
- Nothing sends without an approval
Automation proposes; a person decides
Rallion ingests hiring, funding, launch and personnel signals and normalises them. A war room turns those signals into evidence-cited opportunities and campaigns that a human approves. Human approval gates all execution.
The engine
Approved campaigns run on an owned task queue and daemon: discover prospects, qualify them with an active-learning model — a Gaussian-process classifier with BALD acquisition over FastEmbed embeddings — draft an opener, sync replies and follow up, pacing itself within activity windows and per-channel caps.
Safety as structure
Every prospect on every channel is a thread with a guarded state machine. Illegal transitions raise, and suppression can halt a thread from any state. It's a monorepo: a Django engine of record and the public rallion.co site.