Machine Learning Engineer Graduate (Conversational AI)

Build next-generation conversational agents for live e-Commerce customer service. TikTok’s team in Seattle is hiring a Machine Learning Engineer Graduate (Conversational AI) for a 2027 start. In this role, you will help create a self-evolving unified agent system that operates across 30+ languages, with the runtime, context and memory, post-training/data flywheels, tools and skills, and evaluation/experimentation designed as one closed loop.

Onsite in Seattle, WA. The base salary range for this position in the selected city is $121,600 - $243,200 annually. Compensation may vary based on factors including qualifications, skills, competencies and experience, and location. Base pay is one part of the Total Package and the role may be eligible for additional discretionary bonuses/incentives and restricted stock units.

What you’ll do

• Own agent runtime and behavior control: orchestrate skills, tools, and context; implement loop control with intervention, progressive disclosure, and behavior-level guardrails.
• Build the production safety layer: add pre-flight budgets and timeout truncation, serve-time gates, shadow or swap-in answer delivery, and safe fallback paths.
• Design context and memory for long multi-turn agents: agentic memory with structured note-taking, context compaction and summarization, context editing and observation masking, plus retrieve-then-load retrieval.
• Drive post-training and the data flywheel: use SFT, DPO, and RL approaches to internalize rules into weights so prompts get shorter, distill into smaller serving models, and turn served conversations into training, evaluation, and retrieval signals.
• Enable tools, skills, and MCP workflows: implement tools-as-APIs and connectors, support skill/tool search over large inventories, and govern skill libraries (description conflict handling, trigger evaluations, cross-skill mis-fire matrices) with on-demand loading.
• Evaluate with rigor and scalability: use LLM-as-judge with human-agreement calibration and statistical rigor (paired comparison, confidence intervals, repeated sampling, passk), including held-out and time-rolling splits with overfitting alarms and cascaded judge panels.
• Build the self-evolving loop: case mining, automatic root-cause analysis, constrained candidate generation, replay verification against frozen regression sets, canary deployment, and flywheel iteration.
• Make changes auditable: maintain a candidate registry with exact runtime read-back, change lineage, and an archive of rejected candidates for sampling in later rounds.
• Run online experimentation and causal readout: shadow/canary/A-B testing, non-inferiority gates, traffic-split health, metric definitions that hold up under scrutiny, and off-policy counterfactual evaluation when live A/B isn’t feasible.

What you bring

• Completing or recently completing a Bachelor’s degree in Computer Science, Electrical Engineering, Mathematics, Statistics, or a related discipline.
• Strong Python plus experience with at least one of: C++ / Go / Rust / Java.
• Solid ML/DL/NLP fundamentals with hands-on experience with LLMs or agents (coursework, research, internship, competition, open-source, or a serious side project).
• Basic statistical literacy: confidence intervals, interpreting p-values, and distinguishing “the number went up” from “the system improved.”
• Ability to read a paper or engineering blog and turn it into working code.

Technology you’ll work with

• Python, C++, Go, Rust, Java
• SFT, DPO, RLHF, RLAIF, RLVR
• vLLM, TensorRT-LLM, MoE, KV
• MCP, A/B
• LLM-as-judge
• ACM-ICPC, Kaggle

Benefits

• Day one access to medical, dental, and vision insurance
• 401(k) savings plan with company match
• Paid parental leave
• Short-term and long-term disability coverage
• Life insurance
• Wellbeing benefits
• 10 paid holidays per year
• 10 paid sick days per year
• 17 days of Paid Personal Time (prorated upon hire with increasing accruals by tenure)

Preferred qualifications

• Built the runtime, not only used an API: an agent loop/harness, memory/context management, a RAG or tool-use agent, or a fine-tuned/post-trained model (research, hobby, competition scale included).
• Hands-on depth in one area such as post-training (SFT/DPO/RLHF/RLAIF/RLVR, reward modeling, reward hacking defenses), agent systems (harness, context engineering, MCP/Skills, sub-agents, tool search), evaluation & experimentation (LLM-as-judge, judge calibration, passk, regression suites, A/B, non-inferiority, off-policy evaluation), self-improving systems (evolutionary search, candidate archives, Pareto selection/credit assignment), inference & serving (vLLM/TensorRT-LLM, MoE, KV/prompt caching, cost engineering), or multilingual NLP.
• Publications (for PhD), strong competition results (ACM-ICPC/Kaggle/ML competitions), or notable open-source contributions.
• Taste for the honest number, including confidence in significance before shipping.
• e-Commerce or multilingual experience is a plus, not required.

Application notes

• 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.
• Candidates can apply to a maximum of two positions; applications are considered in the order you apply.
• Applications are reviewed on a rolling basis, and applying early is encouraged.
• TikTok is committed to providing reasonable accommodations in recruitment for candidates with disabilities, pregnancy, sincerely held religious beliefs, or other reasons protected by applicable laws. If you need assistance, reach out at https://tinyurl.com/RA-request.