Resect AI (Resect Artificial Intelligence)
Category: AI Infrastructure
Resect AI builds an "accountability layer" for AI — patented in-stream technology that observes inside LLMs in real time to detect and intercept hallucinations before they reach the user. Resect AI (Resect Artificial Intelligence) was founded in 2019. The company is led by Kevin Owens. Based in Washougal, Washington, USA. Team size: 11-50. Total funding raised: $25M. Latest round: Private Equity. Key investors include Private equity investors (not publicly identified).
- Founded
- 2019
- Headquarters
- Washougal, Washington, USA
- Team size
- 11-50
- Total funding
- $25M
Value proposition
Standard AI observability tools only evaluate model outputs after the fact (post-hoc "black box" monitoring). Resect claims to operate in-stream, looking inside large language models during generation to observe, detect, interpret, audit, and modify model behavior before a hallucination is produced — providing a polygraph/audit layer for neural networks and AI accountability for enterprises facing regulatory compliance.
Products and solutions
Resect NeuroWave Product Suite (enterprise AI observability/interpretability platform): zero-touch observability, global interpretability (identify/neutralize fabrications & hallucinations), real-time AI Audit Log, Agentic Accountability (defense against rogue agent behavior), vLLM integration to detect/mitigate hallucinations at the inference endpoint. Open-source Veritas fact-checkers (0.6B and 8B parameter, Apache 2.0) released on Hugging Face. Built-in foundation model trained via proprietary reinforcement-training methodology.
Unique value
Differentiated in-stream/intervention approach: rather than scoring finished text after generation (like competitors), Resect claims to observe and modify model behavior inside inference in real time to prevent hallucinations before they reach the user — described as a "polygraph for large neural networks." Also open-source, patented technology applicable to major open models (DeepSeek, Qwen, Llama).
Target customer
Enterprises deploying LLMs/AI agents in governed environments where trust, transparency, auditability, and regulatory compliance matter (e.g. financial services, healthcare, media, government).
Industries served
Enterprise AI governance and observability, served verticals span financial markets, healthcare, media, government/compliance, and general enterprise AI.
Technology advantage
Patented in-stream (runtime/generation-time) hallucination detection and behavioral modification; proprietary reinforcement-learning and fine-tuning algorithms that improve factual grounding when applied to open-source models (DeepSeek, Qwen, Llama); self-built LLM giving full control over data, training, and behavior; open-source release (Veritas fact-checkers) plus enterprise suite; vLLM inference integration; a U.S.-based AI safety/Accountability specialists.
How they differentiate
Competitors (Galileo, Patronus, Vectara, etc.) largely score/verify model output after generation (post-hoc fact-checking). Resect differentiates by claiming to look inside the model during generation (in-stream), detecting and modifying behavior prior to output — plus releasing open-source Veritas fact-checkers on Apache 2.0.
Main competitors
Galileo (Galileo.ai), Patronus AI, Vectara (HHEM). Other adjacent: Guardrails AI, Giskard, Braintrust, Arthur AI, Arize Phoenix.
Key partnerships
vLLM integration partnership for inference-time detection, open collaboration with Hugging Face ecosystem via released Veritas model collection, open-source community on GitHub. (Named enterprise partners/customers not publicly disclosed.)
Notable customers
Not publicly disclosed, no enterprise customers named at launch.
Major milestones
Sep 3, 2026: Exited stealth with $25M from private equity investors, launched NeuroWave enterprise suite (waitlist/demo), released open-source Veritas fact-checkers (0.6B & 8B) on Hugging Face (Apache 2.0; models updated ~May 2026).
Growth metrics
30-person workforce (GeekWire, Sept 2026); team of 30 incl. 4 at Washougal HQ. Seed-stage with open-source Veritas fact-checker releases. Revenue/customers not disclosed.
Market positioning
Early-stage stealth-exit (Sept 2026) startup entering the enterprise AI observability / hallucination-prevention &AI accountability layer, well-funded ($25M) and differentiated on in-stream intervention, but its in-stream claim is based on self-reported fact-checking benchmarks (e.g., 72.30% balanced accail) and has not yet delivered independent production benchmarks or broader commercial availability.
Geographic focus
US-centric, HQ in Washougal, Washington (south of Seattle, near Portland, Oregon), 30-person workforce distributed across Seattle, California, New York, and Texas.
Patents and IP
Company claims patented (in-stream) in-stream hallucination detection technology ("Resect says its patented technology operates in-stream"). Veritas fact-checker models (0.6B and 8B) released Apache 2.0 licensed on HuggingFace.
About Kevin Owens
Kevin Owens is Founder, CEO and Chairman of Resect AI. Previously worked with co-founders Tyler Gerber and Tommy Lofgren on Audiience, an advertising and data-science venture focused on cookieless advertising (RuntimeWire, Sep 2026; resect.ai Audiience badge).
Latest news about Resect AI (Resect Artificial Intelligence)
More AI Infrastructure companies
Official website: https://resect.ai
