The role
From TikTok's own posting.
We build the recommendation systems behind TikTok Shop’s shoppable videos and LIVE streams in fast-growing markets across Latin America, Canada, Asia, Australia and the new countries we launch each year. AI and LLMs are transforming e-commerce, recommender systems, and the way research scientists work, and they are opening up new problems and opportunities at the frontier.
LLMs and agents are core to how we work, not just tools we reach for. Our autonomous research agents already run experiments, tune models and diagnose issues in production, and they drive a remarkable share of the improvements we ship.
As more of the execution is automated, what matters most is knowing what to solve. Beyond building and scaling generative models, we own business outcomes end to end: we find the problems, set the direction and define the goals.
We are looking for talented individuals to join our team. As a graduate, you will get opportunities to pursue bold ideas, tackle complex challenges, and unlock limitless growth.
Successful candidates must be able to commit to an onboarding date by the end of the year. Please state your availability and graduation date clearly in your resume.
Responsibilities:
- Build e-commerce recommendation at global scale, end to end. Design and improve the recommendation systems for a platform serving over a billion users, across shoppable videos, LIVE streams and products. You will work on every stage, from retrieval and pre-ranking to ranking and final blending.
- Pioneer generative and LLM-powered recommendation. Use LLMs and generative models to change how recommendation works: generative retrieval, LLM reasoning about user intent (LLM4Rec), and multimodal understanding of AI-generated content (AIGC).
- Optimize decisions with RL, causal inference and operations research. Frame traffic allocation, buyer acquisition and compute allocation as rigorous optimization problems. Use reinforcement learning, causal and uplift modeling, and constrained optimization to decide where each impression and each unit of compute creates the most incremental value.
- Build agents that do research. Create autonomous research agents that mine data for insights, search across competing objectives to push the Pareto frontier, run and analyze experiments, and diagnose production issues. These agents turn weeks of iteration into hours and change how applied science gets done.
- Model the full lifetime of a shopper. Look beyond the next click: model long-term value, repurchase and retention, purchase attribution across surfaces and over time, content fatigue, new-user recommendation and interest exploration. The goal is a system that builds lasting relationships with users, not one-off transactions.
- Turn research into impact. Own your ideas from first prototype to production, validate them through large-scale online experiments, and share what you learn through top-tier publications.
Minimum Qualifications:
- Individuals who are completing or have recently completed a PhD degree in Computer Science, Electrical Engineering, Mathematics, Statistics or a related discipline.
- Strong foundations in machine learning and its underlying math (probability, statistics, optimization), along with solid algorithms and data structures.
- Excellent coding skills in Python or C++ and hands-on experience with deep learning frameworks such as PyTorch.
- Research or project experience in at least one of: recommendation, search or ads; LLMs and generative models; reinforcement learning; causal inference; operations research and optimization; multimodal learning; or AI agents.
- A proven ability to take an open-ended, loosely defined problem, frame it precisely, and drive it to a measurable result.
Preferred Qualifications:
- First-author publications at top venues (NeurIPS, ICML, ICLR, KDD, SIGIR, RecSys, WWW, WSDM, ACL, CVPR, EC) or impactful open-source work.
- Experience with generative recommendation, LLM4Rec, or LLM post-training (SFT, DPO, GRPO, RL-based fine-tuning).
- Experience building LLM agents or automating research workflows, and heavy use of AI coding tools.
- Background in reinforcement learning or bandits, causal and uplift modeling, constrained optimization
- Experience modeling long-term user value, such as retention, user lifetime value or sequential user behavior.
- Experience shipping ML models to production or running large-scale online experiments, for example through internships at consumer-scale products.
- Outstanding results in competitions such as ICPC, Kaggle or KDD Cup