# Alexandr Wang

> 1997– · Entrepreneur; Meta Chief AI Officer; Scale AI founder and director
>
> **Recorded contribution:** Founded Scale AI; built data-labeling infrastructure for AI training; became Meta Chief AI Officer in 2025

## How to use this dossier

Read for a causal chain, not a hero story: inherited problem → contribution → mechanism → downstream capability → limit. Then close the page and complete the reconstruction exercise from memory.

## 1. Historical orientation

American entrepreneur Alexandr Wang co-founded Scale AI in 2016 and served as its CEO until June 2025, when he joined Meta to lead its AI work; Scale appointed Jason Droege interim CEO, while Wang remained a Scale director. Scale built data-labeling, evaluation, and model-support services by combining software workflows with a large distributed labor and contractor base. Wang's place in computing history is infrastructure and institution-building around AI data, not invention of supervised learning or labeling. Scale's platform turns human judgment into a managed production pipeline: a task ontology defines what counts, software routes cases, quality systems compare workers or gold items, and feedback changes instructions. The resulting labels are not raw facts but decisions produced by this sociotechnical process.

## 2. The problem inherited

Machine-learning teams needed large quantities of consistent labeled and evaluated data, but raw human annotation was difficult to specify, route, quality-control, audit, and integrate with model iteration.

## 3. The central contribution

Wang co-founded and scaled an industrial platform for managing human and automated data annotation and AI evaluation workflows.

## 4. Reconstruct the mechanism

1. Translate a model task into labeling instructions, examples, ontology, and quality thresholds.
2. Route items to qualified workers or automated pre-labeling pipelines.
3. Measure agreement and gold-task performance, adjudicate uncertainty, and version the labels.
4. Feed errors into model training and evaluation while tracking provenance, access, cost, and worker impact.

## 5. What changed downstream

- Scale AI became a major supplier in the data pipeline behind autonomous vehicles and generative models.
- Its prominence made hidden annotation labor, data provenance, government contracting, and evaluation power more visible.
- The company's scale made annotation operations and model evaluation strategic infrastructure for AI laboratories, autonomous systems, enterprises, and governments.

## 6. Attribution, limits, and uncertainty

- The company is co-founded and team-built, and model capability belongs to customer and research lineages as well as data infrastructure.
- Label quality is not guaranteed by scale; worker pay, conditions, trauma exposure, privacy, security, ontology bias, and conflicts of interest require independent scrutiny.
- Customer concentration and Wang's 2025 move to Meta create potential conflicts that historical accounts should date explicitly; neither organizational scale nor a prestigious client independently validates label quality.

## 7. Reconstruction lab

Design a 100-item labeling project with ambiguous cases, gold checks, pay assumptions, appeal, and provenance. Report annotator disagreement instead of forcing consensus everywhere. Calculate effective hourly pay and disagreement by case type, then let workers appeal ambiguous gold labels and measure how the ontology changes.

## 8. Evidence trail

- [Founder Alexandr Wang joins Meta; Jason Droege appointed interim CEO](https://scale.com/blog/scale-ai-announces-next-phase-of-company-evolution) — Scale AI
- [Meta's approach to AI safety](https://about.fb.com/news/2025/10/teen-ai-safety-approach/) — Meta
- [NIST AI Risk Management Framework](https://www.nist.gov/itl/ai-risk-management-framework) — National Institute of Standards and Technology

---

*Research checked 2026-08-09. Dates, roles, and claims about living people are historical snapshots. Linked sources remain the authority; this dossier is original instructional synthesis.*