Skip to content

Yrikka

A defense-focused AI startup providing automated red-teaming and simulation tools to stress-test vision and autonomy models before deployment.

Visit Website ↗ + Add to Compare Claim This Company
57/100Incremental Innovator

Overview

Yrikka builds tools for automated, contextual red-teaming and simulation of vision and autonomy AI models. Its APEX API runs adversarial probing and evaluation of exploits against computer vision models, its ORION product generates and adapts training data and finds the conditions under which a vision model fails, and its ATLAS product simulates mission environments — generating a mission world from a text description, map, or photo, then letting a team run an autonomy stack inside it under varying conditions to observe how the system responds.

The San Francisco-based company, founded by CEO Dr. Kia Khezeli and CTO Dr. John Kalantari, serves autonomy builders, defense primes, and government agencies, and its R&D has been funded through contracts with the U.S. Army, U.S. Air Force, and U.S. Special Operations Command (USSOCOM), reportedly totaling more than $2 million. That government contract revenue is a more concrete traction signal than the vendor-reported metrics common among peers at this stage, though it reflects a narrow, defense-oriented customer base rather than broad commercial adoption.

Innovation Matrix Assessment

Innovation Velocity 6/10

Shipped a public API in 2025 and has since expanded to two additional products (ORION, ATLAS) while securing multiple defense contracts within about two years.

Operational Value 6/10

Gives defense and autonomy teams a structured way to stress-test vision and autonomy models before fielding them, addressing a real pre-deployment risk gap.

Market Momentum 5/10

More than $2M in U.S. Army, Air Force, and USSOCOM contract revenue is a stronger, third-party-validated momentum signal than typical pre-seed vendor claims, though the customer base remains narrow.

Category Disruption 5/10

Focused specifically on contextual red-teaming of vision and autonomy models, an underserved niche relative to general-purpose LLM red-teaming tools, but a narrow rather than category-redefining differentiation.

Real-World Efficacy 5/10

Real government contract spend suggests some validated operational use, though no public benchmark results or named incident outcomes were found.

Enduring Relevance 7/10

As defense and autonomy systems increasingly rely on vision AI, rigorous pre-deployment red-teaming is likely to remain a durable, important niche.

Why CISOs Should Care

Gives teams fielding vision or autonomy AI systems a way to find failure conditions and adversarial weaknesses before deployment rather than after an incident.

What Makes It Different

Specializes in contextual red-teaming and mission simulation for vision and autonomy models specifically, rather than general-purpose language model red-teaming.

The Matrix Verdict

57/100 — INCREMENTAL INNOVATOR

A small but credibly funded defense-AI vendor with real government contract traction; an Incremental Innovator whose durability depends on expanding beyond its current narrow government customer base.

Editorial Note: Claims vs. Verified Findings

Contract value figures vary slightly across sources ($1.9M to over $2M); product capability claims come from the company's own site and are not independently benchmarked.

Sources