Software Engineer - Recommendation Infrastructure, Agentic Engineering

About The Team: The Recommendation System Infrastructure team is responsible for building and evolving the large-scale online serving and data infrastructure that powers TikTok's recommendation products globally.

Our mission is to deliver highly efficient, reliable, observable, and scalable infrastructure for recommendation systems. The team works closely with recommendation algorithm teams to accelerate strategy iteration, improve compute efficiency, optimize serving cost, and enable the next generation of AI-native and agentic engineering workflows.

We focus on core infrastructure challenges across online/nearline/offline modules on GPU/CPU, high-performance computing, data pipelines, observability, automation, system reliability, and cost optimization. Our systems are primarily built in C++, while broader infrastructure and automation work may also involve offline data processing frameworks such as Flink, Spark, or other large-scale data systems. A key direction of the team is to build 24/7 closed-loop agentic systems that can observe, diagnose, plan, execute, verify, and continuously improve recommendation infrastructure and iteration workflows.

Responsibilities:

• Improve recommendation system iteration efficiency by enhancing observability, debuggability, automation, scheduling, and execution reliability, with clear impact on cost, iteration speed, reliability, and serving performance.
• Design and build AI-native agentic engineering systems that automate end-to-end engineering workflows, including system observation, issue diagnosis, root-cause analysis, optimization planning, execution, validation, rollout, and post-change verification.
• Build 24/7 closed-loop automation capabilities to replace repetitive, low-efficiency manual engineering work and reduce human overhead in system iteration, performance optimization, and reliability improvement.
• Partner closely with recommendation algorithm teams to support strategy scale-up, improve compute efficiency, accelerate recommendation system iteration, and co-design efficient agentic system infrastructure.
• Optimize high-performance computing frameworks, storage systems, distributed infrastructure components, and runtime systems for recommendation workloads. Minimum Qualifications:
• Bachelor's degree or above in Computer Science, Software Engineering, or a related technical field.
• Experience in building scalable backend systems, distributed systems, infrastructure systems, or high-performance online services.
• Strong programming skills in at least one systems programming language, such as C++, C, Go, or Java.
• Solid understanding of data structures, algorithms, operating systems, networking, and distributed system fundamentals.
• Experience with performance analysis, system debugging, reliability improvement, or large-scale service optimization.
• Strong ownership, problem-solving ability, and communication skills.
• Ability to work effectively with cross-functional teams, including infrastructure teams, recommendation algorithm teams, and product/business-facing engineering teams.

Preferred Qualifications:

• Experience with infrastructure for recommendation systems, search engines, advertising systems, machine learning systems, or large-scale online serving systems.
• Experience optimizing high-throughput, low-latency C++ services in production environments.
• Familiarity with profiling, benchmarking, performance tuning, capacity planning, resource efficiency improvement, and cost optimization.
• Experience with large-scale data processing systems such as Flink, Spark, Kafka, or similar frameworks.
• Experience building end-to-end automation systems based on AI agents, LLMs, workflow orchestration, or closed-loop engineering automation.
• Experience designing agentic workflows for system diagnosis, performance optimization, reliability improvement, change validation, or automatic execution.
• Experience with real-time data pipelines, online training, feature engineering, candidate generation, or recommendation system iteration workflows.