AI Agent Algorithm Engineer Graduate - Global E-Commerce Conversational AI

Overview

Join TikTok's team to build the next-generation unified Agent system that supports global e-commerce customer service in over 30 languages. This innovative platform aims to create a self-improving Agent by integrating training, memory, context management, and evaluation within one closed feedback loop. You will work on the core agent runtime, developing features like agent harnesses and not just building prompts on third-party APIs. As a fresh graduate, you can expect to take ownership of a complete end-to-end component and deliver it to production.

Role Responsibilities

• Develop and enhance the agent runtime which powers customer service agents by managing skills, tools, context, loop control, behavior guardrails, and intervention strategies.
• Design and implement context and memory management for prolonged multi-turn conversations, including structured note-taking, context summarization, editing, masking, and just-in-time retrieval mechanisms.
• Perform post-training and fine-tuning of large language models (through methods like SFT, DPO, RL) and build systems that convert live conversations into training and evaluation signals.
• Design and integrate tools, skills, and multi-channel processor connectors as APIs, along with efficient tool/skill search functionalities across extensive tool inventories.
• Create evaluation frameworks employing LLM-as-judge models calibrated with human agreement, supporting regression, safety, cost, latency-aware testing and closing the gap between offline and online performance.
• Build the self-evolving system loop including replay mechanisms, user simulation, automated root cause analysis, goal setting, and policy updates to facilitate continuous improvement.
• Own a critical end-to-end product surface that supports 30+ languages, driving real business outcomes such as customer satisfaction and issue resolution metrics.

Candidate Profile

• Currently completing or recently graduated with a Bachelor's or Master's degree in Computer Science, Artificial Intelligence, Mathematics, or other quantitative disciplines.
• Proficient in Python and at least one of the following languages: C++, Go, Rust, or Java.
• Strong foundational knowledge in machine learning, deep learning, and natural language processing, with authentic hands-on experience working with large language models or agent systems through coursework, research, internship, competitions, open source, or independent projects.
• Ability to comprehend technical papers or engineering blogs and translate concepts into practical, working code.

Preferred Experience

• Experience building runtime systems (agent loops, harnesses, memory/context management, retrieval-augmented generation, or fine-tuned models) beyond using APIs—even at research, hobbyist, or competition level.
• Hands-on exposure to techniques such as supervised fine-tuning, direct preference optimization, reinforcement learning for human feedback, reward modeling or agent system design involving skills management, context engineering, multi-channel processors, sub-agents, or tool search.
• Experience in evaluation methodologies using LLM as a judge or related benchmarking suites; inference and serving knowledge including tools like vLLM, TensorRT, mixture of experts, and prompt caching; or familiarity with multilingual NLP systems.

About TikTok

TikTok is a leading platform for short-form mobile videos with a mission to inspire creativity and bring joy globally. Headquartered in Los Angeles and Singapore with offices worldwide, the company encourages innovation, resilience, and teamwork through an inclusive culture valuing diverse perspectives.

Diversity and Inclusion

TikTok is dedicated to fostering an inclusive workplace where employees’ unique skills and cultures are respected. The company celebrates diversity to reflect its global community and mission.

Additional Notes

Applicants must be able to commit to onboarding by the end of the year and should clearly specify their availability and graduation dates. Candidates are limited to applying for a maximum of two positions within the company and its affiliates worldwide, with applications evaluated on a rolling basis.