EZKL
EZKL (Zkonduit Inc.) builds zero-knowledge proof infrastructure that converts ONNX-format AI/ML models into verifiable SNARK circuits, serving Web3 developers, DeFi protocols, and AI practitioners via an open-source library and the Lilith cloud platform.
- Company typePrivate
- Founded2022
- HeadquartersChestnut Hill, United States
- Headcount1–10
- GTM typeB2B
- OfferingSoftware
What EZKL does
EZKL (legal entity Zkonduit Inc.) is a cryptography research and development company founded in August 2022 in Chestnut Hill, Massachusetts, that builds zero-knowledge proof infrastructure for verifying AI and machine learning computations. Its core product, the ezkl library, automatically converts ONNX-format machine learning models — sourced from PyTorch, TensorFlow, JAX, scikit-learn, xgboost, and LightGBM — into Halo2-KZG zkSNARK circuits, producing a Solidity verifier contract that can be deployed to any EVM-compatible chain. The system supports tree-based models, transformers, CNNs, RNNs, and custom architectures, and ships GPU acceleration via Ingonyama's Icicle library and Apple Metal bindings, with benchmark results showing ~65.88x faster proving than RISC Zero and ~2.92x faster than Orion.
The company operates a hybrid open-core business model. The ezkl library is free for local developer use, but commercial licensing is required for production deployments past an unspecified version. Lilith, a high-performance compute cluster and orchestration platform, monetizes through a freemium public cluster (with usage caps), an enterprise cluster (unlimited, multi-year contracts), private deployments, and professional services for specialized hardware adaptations. Go-to-market combines product-led developer adoption (pip, npm, cargo, curl distribution, Discord, Telegram, GitHub, Colab notebooks) with enterprise field sales for Lilith cluster access and commercial licensing. The primary customer segments are Web3/DeFi protocol teams (Optimism, Balancer, QuantAMM), AI/ML practitioners verifying model integrity, and emerging use cases in governance, healthcare, and gaming. The company is private, employs ~9 people, and has not publicly disclosed funding rounds.
EZKL firmographics
Firmographics- Name
- EZKL
- Legal name
- Zkonduit Inc.
- Website
- https://ezkl.xyz
- Company type
- Private
- Founded year
- 2022
- Operating status
- Operating
- Headcount range
- 1–10 employees
- Short description
- EZKL (Zkonduit Inc.) builds zero-knowledge proof infrastructure that converts ONNX-format AI/ML models into verifiable SNARK circuits, serving Web3 developers, DeFi protocols, and AI practitioners via an open-source library and the Lilith cloud platform.
- Ownership category
- akta.pro rank
EZKL industry classification
Industry- Product category
- Zero-Knowledge Machine Learning (ZKML) / Verifiable AI Infrastructure
- NAICS
- Custom Computer Programming Services (541511)
- SIC
- Services-Computer Programming Services (7371)
- akta.pro primary industry
- Smart Contract Frameworks & SDKs (developer kits, templates, standards) (FSAPABAB)
- akta.pro secondary industry
- On-Device Inference Runtimes & SDKs (mobile/embedded) (HDAAAJAB)
Keywords
Where EZKL is headquartered
LocationHeadquarters
- HQ city
- Chestnut Hill
- HQ country
- United States
- HQ region
- North America
Offices1 record
Markets served
EZKL business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Technology or R&D, Personnel, Infrastructure, Marketing or Sales, Operations
Revenue model
- EZKL Library Licensing: The ezkl library is free for local installation and use for individuals and teams exploring compatibility or developer experience. However, past version X.X.X, ezkl is not open for commercial usage without proper licensing. This creates a commercial licensing revenue stream for production enterprise deployments.
- Lilith Cloud Platform: High-performance compute cluster for generating EZKL proofs at scale. Offers public cluster (free with usage caps), enterprise cluster (paid, unlimited), and private deployment options. Provides remote orchestration via CLI and REST API.
- Professional Services: Can accommodate specialized hardware requirements and rapidly adapt to new environments. Teams can reach out for custom deployments and specialized setups.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Freemium | Pay-as-you-go | Public Cluster - Free tier for Lilith cloud platform |
| Subscription | Multi-year contract | Enterprise Cluster - Unlimited commercial usage |
| Subscription | Multi-year contract | Private Deployment - Dedicated infrastructure |
| Subscription | Multi-year contract | Commercial Licensing - Library usage for production |
Go-to-market motion3 records
Distribution channels6 records
Marketing channels10 records
EZKL product offering
Product offeringCore offering
EZKL develops and distributes the ezkl open-source library that automatically converts ONNX-format machine learning models (PyTorch, TensorFlow, sklearn, XGBoost, LightGBM) into Halo2-KZG zero-knowledge SNARK circuits, enabling cryptographic proof of correct AI model execution. The company also operates Lilith, a managed cloud compute and orchestration platform that generates EZKL proofs at scale (up to 300,000 proofs per day) on specialized non-cloud hardware, offered as a public free-tier cluster and an enterprise unlimited cluster.
Differentiator
Problem solved
Functional benefit
Brands
- Lilith: A high-performance compute cluster and orchestration platform dedicated to generating EZKL proofs at scale, offering enhanced computational power, reduced latency, and concurrent processing capabilities.
Products and services
- ezkl library A free, open-source developer library that converts ONNX-format machine learning models (from PyTorch, TensorFlow, sklearn, xgboost, LightGBM, and other frameworks) into Halo2-KZG zero-knowledge SNARK circuits. Enables proof of correct model execution without revealing model weights or input data; ships with Python, JavaScript, Rust, and CLI bindings and supports pure software, Intel TEE, and NVIDIA CUDA verification paths. Requires commercial licensing past a specified version for production use.
- Lilith Orchestrator A managed cloud compute and orchestration platform for generating EZKL proofs at scale on specialized non-cloud hardware, capable of up to 300,000 proofs per day. Exposes a CLI mirroring ezkl commands and a REST API for programmatic job triggering, scheduling, and monitoring. Offered as a public cluster (free with usage caps), enterprise cluster (paid, unlimited), and private deployment option for production and miner/validator use cases.
Quantifiable outcome
- ~65.88x faster proving time vs RISC Zero and ~2.92x faster vs Orion
- +5 more outcomes
Companies that use EZKL
Customer profileNamed customers8 records
Segments5 records
Ideal customer profiles4 records
EZKL technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
Integration11 records
AI capability10 records
Feature9 records
EZKL partnerships and signals
Strategic signalPartnerships
14 partnerships are on record, tiered flagship, core and minor.
- OptimismflagshipPartnership to generate ZK proofs for Retroactive Public Goods Funding (RetroPGF) initiative. EZKL provides cryptographic verification that votes were legitimate and allocation algorithm ran correctly while maintaining vote privacy. Piloted with Round 3 allocation worth ~30MM OP.
- IngonyamacoreIntegration of Icicle GPU acceleration library into EZKL engine. Enables ~98% reduction in MSM times and ~35% reduction in total aggregate proof time via CUDA-enabled GPUs.
- LilithcoreHigh-performance compute cluster for generating EZKL proofs at scale. Provides enhanced computational power, reduced latency, concurrent processing, and seamless integration with EZKL CLI.
- QuantAMMflagshipQuantAMM uses EZKL to verify complex quantitative strategies for dynamic AMM pools while protecting proprietary algorithms from alpha leak. QuantAMM is launch partner for Balancer V3.
- BalancercoreQuantAMM deploying dynamic weighted pools directly on Balancer vault using EZKL for strategy verification.
- 0xPARC (Zupass)coreBuilt set membership PCD system using EZKL and Zupass. Demonstrates EZKL integration with PCD SDK for proving group membership without revealing identity.
- Inference LabscoreInference Labs built Doom RL agent verification using EZKL. Demonstrates real-time ZK verification of AI agent inference for gaming applications.
- Trail of BitscoreSecurity audit firm that audited EZKL library, confirming its security for production adversarial environments.
- MicrosoftminorTrusted partner logo displayed on website.
- MITminorTrusted partner logo displayed on website.
- CircleTrusted partner logo displayed on website.
- OpenGradientcorePartnership for scaling trustworthy decentralized computational proofs. OpenGradient x EZKL collaboration for AI inference verification.
- RISC ZerominorCompetitor framework used in benchmarking comparisons. EZKL demonstrated 65.88x faster proving times.
- Giza (Orion)minorCompetitor framework used in benchmarking comparisons. EZKL demonstrated 2.92x faster proving times and 63.95% less memory usage.
Scale indicators6 records
Recent moves6 records
Expansion highlights6 records
EZKL competitors and assessment
Company assessmentEmerging players
- OpenGradient: OpenGradient is a strategic partnership partner of EZKL building decentralized trustworthy compute, including verifiable AI inference. It is a direct ecosystem peer using EZKL's primitives to serve AI verification customers.
- Boundless: Boundless is RISC Zero's general-purpose proof network/marketplace for verifiable compute. It overlaps with EZKL's Lilith as competing managed proof-orchestration infrastructure targeting enterprise verifiable compute workloads.
- Lagrange Labs: Lagrange provides ZK proving infrastructure and coprocessors for off-chain compute verifiability, including ML model inference. It is an emerging player in verifiable compute that targets similar Web3 and enterprise customers.
- Axiom: Axiom builds a ZK coprocessor for Ethereum that enables verifiable on-chain computation across historical state. Overlaps with EZKL where verifiable compute intersects with on-chain ML/analytics, though Axiom is currently broader than ZKML.
Direct peers
- Modulus Labs: Modulus Labs builds on-chain AI and verifiable inference products for blockchain applications, directly overlapping EZKL's on-chain ZKML positioning. Both target the same Web3 developer audience with verifiable ML primitives.
- Giza (Orion): Giza develops the Orion ZKML framework, the closest direct competitor to EZKL in the verifiable machine learning category. EZKL benchmarks claim 2.92x faster proving and 63.95% less memory than Orion across tested model types.
- RISC Zero: RISC Zero builds a general-purpose zkVM (RISC-V based) that supports verifiable compute including ML workloads. It is EZKL's primary benchmarked competitor; EZKL's own data claims 65.88x faster proving times vs RISC Zero for ML model use cases.
Broad incumbents
- Aleo: Aleo is a privacy-focused L1 using ZK cryptography, building developer tooling that includes verifiable computation primitives. It is a broad incumbent in the ZK ecosystem that could expand into verifiable AI use cases analogous to EZKL's.
- Polygon zkEVM: Polygon operates zkEVM and other ZK-based scaling products under a much larger well-funded organization. While not focused on ML verifiability today, it has the resources to incorporate or compete with ZKML primitives at the L2 layer.
Others
- Ingonyama: Ingonyama provides the Icicle GPU acceleration library that EZKL integrates with for hardware-accelerated proof generation. It is an enabling infrastructure/technology partner rather than a competitor, supplying the GPU primitives underlying EZKL's performance claims.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat5 records
Key risks5 records
Key highlights7 records
Customer concentration
EZKL social profiles
Digital presenceEZKL compliance and trust
Trust signalCompliance1 record
EZKL financial estimates
Financial estimateRevenue estimate
Valuation estimate
EZKL leadership team
Management profileNumber of profiles
Profiles2 records
EZKL funding detail
Funding detailFunding overview
Funding rounds1 record
Investors1 record
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
EZKL M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about EZKL
What does EZKL do?
EZKL develops and distributes the ezkl open-source library that automatically converts ONNX-format machine learning models (PyTorch, TensorFlow, sklearn, XGBoost, LightGBM) into Halo2-KZG zero-knowledge SNARK circuits, enabling cryptographic proof of correct AI model execution. The company also operates Lilith, a managed cloud compute and orchestration platform that generates EZKL proofs at scale (up to 300,000 proofs per day) on specialized non-cloud hardware, offered as a public free-tier cluster and an enterprise unlimited cluster.
Is EZKL a public or private company?
EZKL is a private company. It is classified as venture growth investor backed and is currently operating.
When was EZKL founded?
EZKL was founded in 2022. It employs 1 to 10 people.
Where is EZKL based?
EZKL is headquartered in Chestnut Hill, United States, in the North America region.
How does EZKL make money?
Three revenue lines are on record. EZKL Library Licensing is the primary driver. The others are lilith Cloud Platform and professional Services.
Who are EZKL's main competitors?
Emerging players on record are OpenGradient, Boundless, Lagrange Labs and Axiom. Direct peers are Modulus Labs, Giza (Orion) and RISC Zero. Broad incumbents are Aleo and Polygon zkEVM. Ingonyama is listed as an others.
Does EZKL have an API?
Yes. Lilith platform exposes a REST API for programmatic proof job orchestration. Developers can programmatically trigger jobs, chain operations (e.g., gen-witness and prove calls together in a single request), and add new data for workers, all in a single API call. The API also supports callback URLs to receive proofs upon job completion without polling. EZKL library itself offers Python and JavaScript bindings (SDK packages) for local use, as well as a CLI tool for both local (ezkl CLI) and remote (archon CLI) usage. Developer documentation is at docs.ezkl.xyz.
What industry is EZKL in?
EZKL's product category is Zero-Knowledge Machine Learning (ZKML) / Verifiable AI Infrastructure. Its primary akta.pro industry code is FSAPABAB, Smart Contract Frameworks & SDKs (developer kits, templates, standards), with a secondary code of HDAAAJAB, On-Device Inference Runtimes & SDKs (mobile/embedded). Its NAICS code is 541511 and its SIC code is 7371.