Job Description

We are the AI Vertical Scenarios Team within ByteDance's Corporate Service System, based in Dubai. Our mission is to set the standard for AI native applications across ByteDance's internal workplace and procurement scenarios, and to drive consistent AI experiences for employees globally. We build flagship AI applications in-house and use a full scope evaluation suite as our measurement foundation to track how well leading models perform on the real, day to day tasks ByteDance employees carry out.

Why join us
- Real scenarios, real users: the evaluations you design and the AI applications you contribute to are grounded in the actual workflows of ByteDance employees worldwide.
- Hands on with frontier AI: work directly with Claude, Doubao Seed 2.1, and internal agent infrastructure and agent-building tools.
- Broad scope: opportunities to prototype AI solutions to manual workflows, and to contribute to strategic AI projects that shape how ByteDance employees experience corporate services globally.

As a Project Intern, you will contribute to impactful short-term projects and gain hands-on experience in a fast-paced, professional environment. This internship offers the opportunity to develop practical skills, apply your knowledge to real-world challenges, and explore your career interests.
Applications are reviewed on a rolling basis, so we encourage you to apply early.

Online Assessment
Candidates who pass resume screening will be invited to participate in Our Company's technical online assessment.

Responsibilities:
- Core evaluation work:
Own a defined scope of ByteDance's AI evaluation benchmark end to end, with ownership to identify problems and propose solutions. You'll run the full research and evaluation loop on some of the world's most widely used AI Agents, surfacing capability gaps and failure modes, and turning those findings into actionable insights that directly drive the Agent's iteration at global scale. This includes:
1. Ground the work in real user needs through user interviews, literature review, and usage data analysis.
2. Propose research methodology: how queries are generated, how rubrics are structured, and how models are scored.
3. Write curated queries with prerequisites, across three difficulty levels: Simple, Medium, and Complex.
4. Write rubrics: checklist items defining a correct answer (e.g., for a leave request: leave type, date parsing, balance check, approver routing, form completeness, submission state).
5. Run models against the queries and score outputs on an ongoing cadence, reviewing multiple agents and models against the benchmark each month.
6. Automate the evaluation so the loop can be re-run as models and internal systems evolve.
7. Write the evaluation report covering findings, failure patterns, and recommendations.

- Additional project work:
Beyond the core evaluation work, interns will have full agency to support and lead ad hoc projects that shape how AI gets implemented and adopted across ByteDance globally, including:
1. Identifying manual workflows across the business and prototyping AI solutions to automate them.
2. Contributing to AI product work by shaping agent capabilities, and helping ship new AI features into internal workplace and knowledge tools.