Graduate Research Scientist, Conversational AI Agent Systems

TikTok; Seattle (WA); On-site; USD 154,000 - 301,000; Full time

Job summary

TikTok is hiring a new graduate to join its AI/ML team in Seattle. You will own an end-to-end surface from day one and ship it to production, tackling a range of challenges from agent runtimes to memory management and evaluation at scale.

You will explore post-training methods and tool governance while contributing to a fast-evolving product used across 30+ languages. The role emphasizes building robust, auditable systems, cross-language support, and visibility into the entire lifecycle of

Qualifications

• PhD or recent grad in CS/AI/Math/Quantitative field.
• Strong Python with one of C++/Go/Rust/Java.
• Solid ML/DL/NLP fundamentals with hands-on experience with LLMs or agents.
• Basic stats literacy—confidence intervals, p-values, and interpretation.

Responsibilities

• Architect and implement end-to-end agent systems and runtimes; manage loop control, guardrails, and safe fallbacks.
• Design and maintain long-term memory and context management for multi-turn agents.
• Apply post-training methods (SFT/DPO/RLHF) to improve prompts and models.
• Develop tools, skill libraries, and governance for large inventories and on-demand loading.
• Perform rigorous evaluation with human alignment, AB testing, and regression checks.
• Build auditable pipelines: candidate registries, run-time history, and change lineage.
• Lead online experiments (shadow/canary/A-B) with robust metric definitions and safeguards.
• Ensure safety and anti-gaming measures in the evaluation and release process.
• Own high-leverage end-to-end surface and ship to production across languages.

Skills

Python

C++/Go/Rust/Java

ML fundamentals

NLP fundamentals

Reading papers into code

Education

PhD in CS/AI/Math/Quantitative field

Job description

TikTok is hiring a new graduate to join its AI/ML team in Seattle. You will own an end-to-end surface from day one and ship it to production, tackling a range of challenges from agent runtimes to memory management and evaluation at scale.

You will explore post-training methods and tool governance while contributing to a fast-evolving product used across 30+ languages. The role emphasizes building robust, auditable systems, cross-language support, and visibility into the entire lifecycle of