Machine Learning Engineer Graduate

About the Role

Global E-Commerce (TikTok Shop) is a fast-growing business within TikTok, driving significant revenue growth. The Global E-Commerce Content Recommendation team is responsible for the end-to-end recommendation stack for e-commerce content (video, image-text) globally. This includes retrieval, ranking, multi-queue blending, the supply ecosystem, cold start problems, and the browsing-to-purchase experience for hundreds of millions of users. The team is rebuilding the recommendation stack on LLM foundations, aiming to create the most advanced recommendation system. As a graduate, you will have opportunities to pursue bold ideas, tackle complex challenges, and unlock growth. Successful candidates must commit to an onboarding date by the end of the year.

Key Responsibilities

• Build and rewrite an industrial recommendation system serving a billion-scale user base across short-video, livestream, and product scenarios, covering retrieval, pre-ranking, ranking, and blending end-to-end.
• Scale recommendation models like LLMs, pushing ranking models from hundreds of millions to billions of parameters.
• Chart the scaling laws of recommendation: behavior-corpus pretraining; multi-scenario, multi-task, multi-stage joint training; ultra-long behavior-sequence modeling (10K+ events) using KV caching, sequence compression, user/generation (U-G) disaggregated serving, speculative decoding, and dynamic batching.
• Build one-stage generative retrieval, reframing retrieval as generation by tokenizing the item space into semantic IDs (RQ-VAE / SID) and training autoregressive models grounded in MLLM semantics.
• Address open problems in item tokenizers, SIDs stability, post-training the generator directly on live user feedback (preference optimization, GRPO-style RL), and decoding under a millisecond budget (beam search, decoding constrained to the valid item space, and test-time scaling).
• Inject world knowledge from large models to mine latent user interests and semantic representations.
• Use reasoning models for explicit chain-of-thought inference over long-horizon user intent.
• Push training and inference to the hardware limit using custom CUDA / Triton fused kernels, memory and computation-graph optimization, distributed training and inference acceleration, mixed precision and low-bit quantization.
• Rewrite R&D with agents for automated feature mining, pipeline generation, experiment configuration, training orchestration, automated evaluation, online-diagnosis attribution, and bad-case mining.
• Do original work on open problems like long-term value modeling, repurchase and retention, transaction attribution, fatigue modeling, new-user recommendation, incremental value modeling, interest exploration, and LLM4Rec.
• Expected to produce internal papers, patents, and publications at top external venues.

Requirements & Tech Stack

• ML and engineering fundamentals: Understanding of model mathematics, clean and efficient code, strong command of algorithms and data structures.
• Deep research or engineering practice in at least one of: LLMs / foundation models, NLP, CV, RL, or recommendation / search / ads.
• Genuine enthusiasm for LLM / LRM techniques for production deployment.
• Strong problem definition and decomposition skills.
• PhD degree in Computer Science, AI, Mathematics, Statistics or a related discipline.

Preferred Qualifications

• Publications at KDD, SIGIR, RecSys, WWW, ACL, NeurIPS, ICML, ICLR, or comparable venues; or high-quality open-source work.
• CUDA / Triton kernel development, source-level deep-learning-framework optimization, large-scale distributed training, or high-performance inference deployment.
• Hands-on experience with LLM post-training (SFT / RLHF / DPO / GRPO), agent-system construction, or inference acceleration.
• Led or deeply contributed to a key project in search, ads, recommendation, or large models, with a complete problem-to-online-impact loop.
• Awards in ACM-ICPC, NOI, Kaggle, or comparable competitions.
• Heavy user of AI coding and agentic workflows for building systems and optimizing models.