Arize AI
Arize AI develops an observability and evaluation platform for AI systems in production, serving enterprises that depend on machine learning and large language models to run business-critical functions.
Profile
Observability, evaluation, and monitoring platform that helps teams deploy and troubleshoot AI models and agents reliably in production.
Arize AI develops an observability and evaluation platform for AI systems in production, serving enterprises that depend on machine learning and large language models to run business-critical functions. Founded in 2020 by Jason Lopatecki and Aparna Dhinakaran, the Berkeley-based company started by addressing model drift and performance degradation in traditional ML pipelines—the "observability gap" in MLOps. As the industry shifted toward generative AI, Arize expanded its platform to provide tracing, evaluation, and monitoring for LLMs and AI agents.
The company operates a two-tier go-to-market: Phoenix (open-source observability framework, 2+ million monthly downloads) and Arize AX (enterprise managed platform). Named customers include Booking.com, DoorDash, Duolingo, Hyatt, PepsiCo, TripAdvisor, Uber, and Wayfair. In February 2025, Arize closed a $70 million Series C led by Adams Street Partners, bringing total funding to $131 million.
Earlier rounds included $38 million Series B (September 2022, TCV) and $19 million Series A (September 2021, Battery Ventures). In March 2025, Arize acquired Velvet, a developer-focused AI evaluation platform, to accelerate product engineering and expand its capabilities. The company also secured a $1.2 million SBIR contract with the Air Force (AFWERX) in July 2025 to advance AI engineering tools for government use.
By April 2026, Arize had grown to approximately 191 employees. The company faces competition from Galileo ($68 million raised) and Patronus AI ($20 million raised) in a crowded but differentiated market where scale and ecosystem integration remain key advantages.
Who buys this
- Enterprises running large-scale recommendation systems and personalization (travel, e-commerce, food delivery)
- Financial services and pricing optimization teams using AI for forecasting and risk
- SaaS companies building conversational AI and LLM-powered features
- Government and defense agencies developing secure AI systems
- Ad tech and content platforms optimizing ranking and ranking models
Publicly disclosed clients
- Booking.com
- DoorDash
- Duolingo
- PepsiCo
- TripAdvisor
- Uber
- Wayfair
- Condé Nast
Strengths and what to watch
Strengths
- Vendor-agnostic, open-source foundation (Phoenix) with strong adoption and developer trust—critical moat in fragmented AI stack
- Enterprise-grade infrastructure managing 1 trillion spans and 50 million evaluations monthly; deep integrations with Azure, AWS, and GCP
- Strategic acquisition of Velvet (March 2025) strengthens developer experience and time-to-value, addressing the largest friction point for enterprise adoption
Watch for
- Customer concentration risk: major customers (Uber, Booking, etc.) represent outsized revenue; churn at single large account could materially impact growth
- Competitive pricing pressure as Galileo and Patronus AI add observability features; differentiation increasingly tied to eval quality and ecosystem reach, not monitoring alone
- Government revenue dependency: AFWERX contract ($1.2M/yr) is meaningful but niche; unclear if military/defense becomes strategic pillar or distraction
Recent moves
Key Information
- Industry
- AI Observability
- Founded
- 2020
- Headquarters
- Berkeley
Frequently Asked Questions
What is Arize AI?
Arize AI is an observability and evaluation platform helping enterprises deploy and troubleshoot AI models and agents in production. Founded in 2020, it serves companies relying on machine learning and large language models for business-critical functions, addressing the observability gap in AI operations.
Does Arize have an open-source option?
Arize offers two tiers: Phoenix, an open-source observability framework with over two million monthly downloads, and Arize AX, an enterprise managed platform. This dual approach builds strong developer trust while serving enterprise customers needing managed infrastructure, dedicated support, and advanced features.
Which companies use Arize AI?
Arize serves major enterprises including Booking.com, DoorDash, Duolingo, Uber, Wayfair, and PepsiCo. Customers span travel, e-commerce, financial services, SaaS, and government sectors. They use Arize for monitoring recommendation systems, personalization, AI forecasting, risk optimization, and LLM-powered features deployed in production environments.
How much data can Arize monitor?
Arize manages enterprise-grade infrastructure handling one trillion spans and fifty million AI evaluations monthly. It integrates deeply with AWS, Azure, and Google Cloud, supporting large-scale deployments including recommendation systems, personalization engines, conversational AI, and other mission-critical production ML workloads.
How does Arize compare to other AI observability platforms?
Arize competes with Galileo and Patronus AI but differentiates through its vendor-agnostic open-source foundation, enterprise-scale infrastructure, and recent acquisition of Velvet to strengthen evaluation capabilities. Scale and ecosystem integration remain key competitive advantages in the growing AI observability market.
How much funding has Arize raised?
Arize closed a $70 million Series C in February 2025 led by Adams Street Partners, bringing total funding to $131 million. Previous rounds included $38 million Series B (2022) and $19 million Series A (2021). By April 2026, the company had grown to approximately 191 employees.
How Arize AI compares
Direct head-to-head against 3 competitors. Picked by 7wData.
Arize AI
- Positioning
- Observability, evaluation, and monitoring platform that helps teams deploy and troubleshoot AI models and agents reliably in production.
- Customer segments
- Enterprises running large-scale recommendation systems and personalization (travel, e-commerce, food delivery)
- Strengths
- Vendor-agnostic, open-source foundation (Phoenix) with strong adoption and developer trust—critical moat in fragmented AI stack
- Watch for
- Customer concentration risk: major customers (Uber, Booking, etc.) represent outsized revenue; churn at single large account could materially impact growth
- Recent moves
- Arize AI selected by AFWERX for $1.2M SBIR contract to enhance Air Force AI systems on NIPRNET
Galileo
- Positioning
- LLM evaluation and agentic observability platform targeting enterprise GenAI teams via proprietary evaluation models and a free developer tier.
- Customer segments
- Fortune 500 enterprise AI teams building GenAI applications; named buyers include Comcast, Twilio, Reddit, and HP.
- Strengths
- Proprietary Luna-2 evaluation model family enabling scalable, low-latency in-production guardrails for multi-agent systems.
- Watch for
- 153-person team versus growing enterprise demand; delivery and support capacity under stress as customer count scales rapidly.
- Recent moves
- Launched free Agent Reliability Platform combining observability, evaluation, and guardrails for multi-agent AI systems, July 2025.
Weights and Biases
- Positioning
- ML experiment tracking platform extended into LLM observability via Weave; now bundled with CoreWeave GPU cloud after $1.7B acquisition.
- Customer segments
- ML engineers and AI researchers at enterprises and labs already using W&B for model training and evaluation; 1,400-plus organizations including Nvidia and AstraZeneca.
- Strengths
- Dominant experiment lineage and artifact versioning embedded in training workflows at major AI labs across 1,400-plus organizations.
- Watch for
- CoreWeave acquisition creates incentive to favor CoreWeave infrastructure customers, eroding vendor-agnostic positioning that attracted many enterprise accounts.
- Recent moves
- Acquired by CoreWeave for $1.7 billion, completed May 2025, integrating W&B developer platform with GPU cloud services.
Patronus AI
- Positioning
- LLM evaluation and hallucination detection platform for enterprises; pivoting toward agent simulation infrastructure targeting foundation model labs.
- Customer segments
- Enterprise AI teams testing LLM safety, hallucination risk, and compliance in regulated-industry GenAI deployments.
- Strengths
- Proprietary Lynx hallucination detection model and Percival agent debugger covering over 20 agentic failure modes out of the box.
- Watch for
- Late-2025 pivot to AGI simulation research and ORSI training methods risks defocusing the enterprise evaluation product core buyers purchased.
- Recent moves
- Introduced Digital World Models, RL Environments, and Open Recursive Self-Improvement method for agent simulation training, late 2025.
Sources
- arize.com — Company overview, products (AX Platform, Alyx, Phoenix OSS), scale (1 trillion spans, 50M evaluations/month)
- arize.com — Founders (Jason Lopatecki, Aparna Dhinakaran), mission, team structure, offices (Berkeley, New York), employee count (43 open positions as of fetch)
- www.prnewswire.com — Series C: $70M, February 2025, lead investor Adams Street Partners, participating investors (M12, SineWave, OMERS, Datadog, PagerDuty), returning investors, named customers (Booking, Condé Nast, Duolingo, Hyatt, PepsiCo, Priceline, TripAdvisor, Uber, Wayfair)
- techcrunch.com — Series C funding, CEO comment on growth, competitive positioning vs Galileo and Patronus, business model, market context
- tracxn.com — Founded 2020 in Berkeley, total funding $131M, Series B $38M September 2022, Series A $19M September 2021, 191 employees April 2026, acquired Velvet March 2025
- arize.com — Series A: $19M, September 28 2021, led by Battery Ventures, founding context and market opportunity
- www.prnewswire.com — Series B: $38M, September 8 2022, led by TCV, prior investors Battery Ventures and Foundation Capital
- siliconangle.com — Velvet acquisition March 2025, strategic rationale (talent, tech complementarity), Chris Hendel as CTO leading platform engineering, market context
- arize.com — Named customers including Booking.com, PepsiCo, Siemens, DoorDash, Instacart, Reddit, Roblox, Adobe, Wayfair, TripAdvisor, Uber; customer testimonials; primary use cases