Software Architect - PNS AI Governance

岗位职责 / Responsibilities

TikTok's AI Governance team is chartered with ensuring responsible development and deployment of AI across the platform - from large-scale training pipelines through production inference. As foundation models and generative AI move deeper into product surfaces, the team owns the controls, policy enforcement, and technical safeguards that keep user data protected and model behavior aligned with regulatory and ethical commitments.

This role, Software Architect - AI Governance, will design and implement scalable, reliable, and high-performance systems that govern how data and models are trained, evaluated, released, and served at TikTok scale. The architect will translate fast-evolving privacy, safety, and regulatory requirements into concrete platform capabilities, and will partner closely with ML platform, research, product, privacy, and security teams to land solutions that work at the scale of TikTok's training and serving infrastructure.

- Design and implement scalable, reliable, high-performance systems for data and model governance across the full ML lifecycle - data ingestion, training, evaluation, deployment, and online inference.

- Architect controls for large-scale model training and deployment: training-data lineage and provenance, dataset approval and taxonomy, model registries, evaluation gates, and safe-release workflows for foundation and generative AI models.

- Develop innovative solutions to tackle emerging risks in Generative AI - including prompt-injection defense, data leakage and memorization mitigation, output safety filtering, agentic-system guardrails, watermarking, and provenance signals.

- Partner with ML platform and research teams to embed governance hooks (policy checks, PII/PNS filtering, red-team evaluations) directly into training and serving paths without regressing model quality or throughput.

- Collaborate with privacy, security, and legal to ensure systems comply with applicable regulations (GDPR, CCPA, EU AI Act, sector-specific rules) and with TikTok's own commitments on responsible AI.

- Drive the technical roadmap for AI governance platform capabilities; identify and evaluate new techniques, tools, and research that strengthen the governance posture at scale.

- Provide technical leadership and mentorship to engineers across the organization on AI governance, data governance, and generative-AI risk topics.

任职要求 / Requirements

Minimum Qualification(s)

- Bachelor's or Master's degree in Computer Science or related field.

- Strong background in large-scale distributed systems, databases, messaging and streaming systems, and caching.

- Hands-on experience with the ML lifecycle at production scale - training pipelines, feature/data stores, model registries, or online serving systems.

- Expertise in data and/or model governance for large-scale model training and deployment: lineage, access control, evaluation, release gating, monitoring.

- Track record of developing innovative technical solutions for emerging risks in Generative AI models (e.g., prompt injection, data leakage / memorization, output safety, alignment evaluation, agentic-system safety).

- Experience with microservices architecture and containerization (Docker, Kubernetes), cloud platforms such as AWS or GCP, and Agile development and DevOps practices.

- Strong communication and collaboration skills; ability to work independently across multiple priorities.

Preferred Qualification(s)

- 5+ years of industry experience in a software development environment, with significant time spent on ML platform, AI safety, or data/AI governance.

- Familiarity with privacy-enhancing technologies (differential privacy, federated learning, secure enclaves, PII detection / de-identification).

- Familiarity with applicable regulations (GDPR, CCPA, EU AI Act) and industry standards (NIST AI RMF, ISO/IEC 42001).

- Experience with foundation-model training, RLHF, evaluation harnesses, or red-teaming frameworks.

- Publications, patents, or open-source contributions in AI governance, AI safety, or responsible AI.