Ilya Kirnos

Founding Partner + CTO @ SignalFire — the technical / engineering brain of the firm.

Snapshot

  • Email: ilya@signalfire.com
  • LinkedIn: /in/imkirnos
  • Role: Founding Partner, Venture + CTO
  • Firm joined: 2013 (co-founder)

Background

  • Google (2004–2012):
    • Tech Lead, Gmail Ads + AdWords Performance and Scalability
    • Performed due diligence on Urchin acquisition (→ became Google Analytics)
    • Founded and led engineering for Google Prediction Markets
  • CardSpring — Led platform backend (acquired by Twitter)
  • SignalFire — Co-founded 2013; built the firm’s Beacon AI product (proprietary data/sourcing engine).

Why his joining matters for init.inc

SignalFire’s structural moat is data (Beacon AI). Ilya is the technical engine behind it. When a deeply-technical engineering-led founding partner joins a meeting, it signals the partnership wants to stress-test the technical thesis before any conviction. Lead with technical depth + AI infra wedge — he’ll dig there.

Fundraise Relationship (init.inc)

  • 2026-05-28 3pm — joining Ryan+Ilya+Sazzad meeting (Ryan+Ilya//Sazzad (Init) Chat).
  • Read: Ryan re-activating the dead-warm thread (ghost since 5/5) AND bringing the Founding-Partner-CTO = real reactivation signal; not a courtesy call. Ilya joining means technical due diligence is the bar — he’s there to stress-test architecture, not the deck.

Pre-meeting tactical brief (2026-05-28 3pm)

Open with: Acknowledge his Beacon AI work directly — “We’ve thought about Beacon as a model for how an AI-first VC differs from a brand-VC adding data overlays; init’s architecture is similarly AI-native vs MSPs bolting AI onto labor-cost org structures.” Signals you’ve done the homework on what he built.

Demo pivot: Skip the deck → walk him through the agent orchestration layer. Specifically:

  • How init’s agents reason over incident → CMDB → resolution paths without requiring a mature ServiceNow CMDB (mid-market reality)
  • Where the agents fail-gracefully + how human-in-the-loop is structured (Leigh-style “agents don’t work reliably” was the consensus 18 months ago — show how init’s narrow vertical scope solves it)
  • The training-loop economics — every resolved incident improves the next agent’s prior; cost-per-incident decreases over time, not the other way around

Anticipated objections + answers:

  • “Is this just a wrapper on LLMs?” → No: the orchestration layer + the proprietary training corpus from operating Delve at scale (real customer incident data) is the moat. LLMs are commoditizing; the labeled-incident dataset isn’t.
  • “Why won’t ServiceNow / incumbents catch up?” → They can add features, but they can’t restructure their ~70% labor cost base into ~10%; they have shareholders and ~$200B market cap to protect. We’re org-architecture-native AI.
  • “What’s your sourcing edge?” → Bottoms-up technical-buyer demand + founder credibility from Delve → init pivot; we’re already winning logos at companies that have ServiceNow contracts.

Closing CTA: Ask Ryan + Ilya: “What does an internal-champion process look like at SignalFire? Want to bring Wayne (CEO) or another partner into a follow-up so we can pressure-test the org-architecture thesis with the full team?”

signalfire · ryan-wexler