First Check

Stage 01

Source

Inbound and outbound signal parsed into verifiable facts. Founders and ideas recognized across opportunities. Data lands in the Card with source attached.

Tech stack

Next.js / Temporal / LlamaParse / Zod / Drizzle ORM / PostgreSQL (RLS) / S3

Theory

Entity resolution & fact provenance
Evidence-cited knowledge capture

fig 1 — many sources, one card

Stage 02

Research

Investment thesis drives evaluation. Due diligence per axis: Founder, Market, Finance, Legal & Ops. Risks and gaps flagged for interview.

Tech stack

Temporal / OpenAI Agents SDK / pgvector / TypeScript calculation engine

Theory

Investment thesis modeling
Zero-knowledge proofs for due-diligence sourcing

fig 2 — thesis driven analysis

Stage 03

Interviewing

Agent-driven interview resolves open questions, converges founder and investor towards optimal deal within pre-approved bounds. Anormaly escalates to the investor, who keeps final authority.

Tech stack

ElevenLabs Agents / WebRTC / Server-Sent Events / Clerk / Temporal

Theory

Multi-user social / personality modeling
Negotiation theory (BATNA, Nash bargaining)

fig 3 — convergence within bounds

Nimbus AI

Opportunity · Seed · Infra / AI Tooling

EMPTY

01 · Source

09:02:11Founders: Ada Chen (ex-Stripe infra) & Raj Mehta (PhD, distributed systems)deck

09:02:47Idea: serverless GPU orchestration — “Vercel for model inference”form

09:03:15Traction: 1.2k GitHub ⭐ in 6 weeks, 14 design partnersgithub

09:03:58Raising $3M seed; warm intro via portfolio foundersocial

02 · Research

PDF

founder_assessment.pdf

lens: FOUNDER · verdict: strong