AIxBlock

AIxBlock is an enterprise AI training data company that supplies speech, audio, and large language model datasets to Fortune 100 teams and regulated industries. The company operates a proprietary self-hosted platform that runs annotation and collection workflows inside the client's own infrastructure, so raw data never transits or rests on AIxBlock servers — a design it calls "True Exclusivity." The result is a managed data partner that satisfies CISO and compliance requirements that typically stall enterprise AI projects for months.

Origins and evolution

The team behind AIxBlock traces its roots to 2019, when the founders began building AI training data for machine learning teams at large enterprises. Over seven years they delivered large-scale data projects for clients including Oracle, AWS, Apple, Uber, Uniphore, and Chainalysis, accumulating more than 200,000 hours of delivered audio and annotation work across 100+ languages.

The company was formally incorporated in 2023, and for a period between roughly 2023 and early 2025 positioned itself as a Solana-based decentralized AI development platform — offering a GPU compute marketplace, a model marketplace, and low-code AI workflow automation, with a Token Generation Event (TGE) announced for Q2 2025. That decentralized infrastructure layer, which promised up to 90% savings on compute costs versus traditional cloud providers and drew a partnership with node management platform Hivello in January 2025, was backed by the European Union's innovation programme.

By 2025 the company had repositioned. The current website carries no mention of Solana, a native token, or a public GPU marketplace. The marketplace pages no longer resolve. AIxBlock now markets itself as a sovereign AI data platform whose competitive differentiation is enterprise specialization, data residency architecture, and a seven-year delivery track record — not blockchain infrastructure. The self-hosted platform still includes a GPU marketplace component, but it is offered as part of a private, on-premises deployment rather than a decentralized public network.

The EU has continued to back the company: in August 2025, AIxBlock received a $1.74 million grant from a European Union innovation programme, bringing total disclosed EU-source funding to that amount.

Core products and services

Speech Data Services — End-to-end collection, transcription, and annotation of voice recordings in 100+ languages and dialects. AIxBlock recruits professional voice talent or natural speakers, handles speaker-level timestamps and verbatim transcription, and applies layered quality controls (consensus scoring, blind testing, tiered QA/QC/QC2 roles) across every project.

Text and Dialogue Data Services — Conversation annotation, intent and entity labeling, RLHF-style preference data (pairwise comparisons for reward model training), and full fine-tuning dataset construction for LLMs. The service targets Foundation Model labs and internal AI product teams building copilots or domain-specific chat systems.

OTS Audio Library — An off-the-shelf catalog of hundreds of thousands of hours of real-world call center recordings, covering US, Indian, and Philippine English accents plus multiple Indian-language variants. Enterprises license this pre-existing audio rather than commissioning fresh collection, shortening time-to-data significantly.

Sound and Environment Audio — Environmental sounds, background noise, machine sounds, and acoustic scene recordings used in audio classification, noise detection, and sound recognition model training.

Self-Hosted Platform — The entire AIxBlock data engine — collection tooling, annotation workflows, model training and fine-tuning environment, and a GPU marketplace — packaged for on-premises or private cloud deployment. The client supplies storage; AIxBlock manages the annotation workforce and QA processes on top. GDPR, HIPAA, and air-gapped deployment configurations are all supported.

Quality and governance architecture

AIxBlock's quality model is built around several layers it describes as proprietary. Contributor onboarding uses KYC with biometric enrollment, device fingerprinting, and credential validation. Active sessions are monitored with random biometric re-authentication and liveness verification. Work quality is measured via a consensus mechanism that compares each contributor's output against a project benchmark and promotes the highest performers into QA and QC roles. Blind test tasks are inserted continuously to catch drift.

The platform also applies behavioral anomaly detection — work-pattern baselines, automation detection, and root-cause analysis on blind-test failures — to filter proxy work and identity mismatches. Local Subject Matter Experts localize guidelines and deliver annotator training in-language rather than relying on translation alone.

For regulated clients, the self-hosted architecture means AIxBlock physically cannot resell proprietary data or use it to train other customers' models, because no copy of the data ever resides on AIxBlock infrastructure.

Security and compliance

The self-hosted deployment ships with SOC 2 and ISO 27001-ready configuration scripts. Data sovereignty is enforced at the architecture level: the client's storage is connected from project kickoff, and all collected or annotated assets flow directly into that storage with no intermediate copy retained. This satisfies the regulatory requirements of banking, healthcare, government, and defense customers who cannot send data to third-party SaaS systems.

Team and backing

Co-founders include Ha Dao (CEO), Tony Dinh (Head of Product), and Jackie Lu. The team built the company without venture capital in its early years, relying primarily on enterprise contracts and EU innovation grants. Institutional backing from the European Union provides credibility on data governance standards relevant to the GDPR-compliant regulated sectors AIxBlock serves.

Fit in the Solana ecosystem

AIxBlock's current relationship to Solana is historical rather than active. The platform was architecturally connected to Solana during its 2023–2024 decentralized-AI phase, when on-chain settlement and tokenized compute access were part of the roadmap. That positioning has since been withdrawn: the TGE was not publicly confirmed as executed, and no Solana-native contracts or token infrastructure appear on the current platform. As of mid-2026, AIxBlock is best understood as an enterprise AI data company with a Solana-origin story rather than an active protocol or DeFi project. Teams looking for decentralized GPU compute or an AI token on Solana should treat the earlier roadmap claims with caution and verify directly with the company before building on that infrastructure.