Job Description
ABOUT CITADEL AI
Citadel AI builds software products to evaluate, monitor, and govern AI systems. Our mission is to make the world’s AI systems more reliable and trustworthy.
We’re located in Tokyo, but we have an international team and global customers. Our goal is to build a world-class startup from Japan, and our team members have previously worked at leading companies such as Google, Paidy, Apple, Stripe, and Meta.
This is a unique opportunity to join a team in Tokyo building new products for a global market. As an early employee at Citadel AI, you’ll have significant impact and ownership over the growth of our startup. We’re supported by Japan’s best investors and a top-tier enterprise customer base, and are currently scaling our business in the 1 → 10 stage.
ABOUT THE ROLE
As an AI Research Engineer at Citadel AI, you will research technologies to ensure the safety and reliability of frontier AI systems, including frontier models, agent harnesses, and multi-agent and multi-modal systems. You will also study the broader impact these technologies may have on society and organizations. This role goes beyond literature reviews and field research: you will also design frameworks, build prototypes, and develop benchmarks.
Your work will not only be compiled into research reports, but may also be integrated into Citadel AI’s product portfolio. This is a key position for creating the foundation of our long-term technical differentiation. Compensation is a flexible combination of salary and stock options.
RESPONSIBILITIES
- Conduct research on safety evaluation methods for frontier models, agent harnesses, multi-agent and multi-modal systems, and related technologies.
- Investigate risk assessment methods, attack methods, defense methods, and design and implement benchmarks.
- Participate in technical discussions and joint research with public institutions and major enterprises.
- Plan and execute R&D projects, and prepare evaluation reports.
- Propose ways to integrate your research into Citadel AI’s product portfolio.
JOB REQUIREMENTS
- 2+ years of experience in research and development in generative AI or equivalent experience.
- Ability to investigate and analyze technical trends from research papers, including arXiv and conference publications, organize findings systematically, and document them clearly.
- Knowledge of LLMs, VLMs, and the AI ecosystem.
- Depending on the team, you must be fluent in either English or Japanese.
NICE TO HAVE
- Publications or presentations at major international conferences such as CVPR, ICML, or NeurIPS.
- Experience making technical proposals and delivering reports to clients in contract R&D projects.
- Experience building and operating AI agents.
----------
〜未踏のAIリスクに挑み、次世代の技術とそのインパクトを調査・研究する〜
役割・ミッション
マルチモーダルやマルチエージェントシステム、独自LLMなど最先端AIの安全性や信頼性を確保するための技術、さらにそれが社会や組織にもたらす影響を研究するのが、今回募集するポジションの役割です。論文調査やフィールドリサーチ、フレームワークの考案だけでなく、プロトタイプの開発やベンチマーク作成までを行い、Citadel AIの技術的優位性と社会的信頼性を確立していただきます。
ここで得られた知見は、調査報告書に留まらず、将来的にはCitadel AIのプロダクトに統合される可能性があり、技術的優位性の源泉を作る重要なポジションです。報酬体系については「給与とストックオプションの組み合わせ」なども相談可能です。
具体的な業務内容
- マルチモーダルAI、マルチエージェントシステム等の安全性評価に関する技術調査
- リスク評価手法、攻撃手法、防御手法の調査やベンチマークの考案と実装
- 公的機関や大手企業との技術ディスカッションおよび共同研究
- 研究開発の計画と遂行、評価レポートの作成
- 獲得した技術知見の自社プロダクトへの技術移転・統合の提案
必須スキル・経験(MUST)
- 生成AI分野における研究開発ないしはそれに類する経験(2年以上目安)
- 最新の論文(arXiv/学会発表等)から技術的な動向を調査・分析し、体系化したうえでドキュメント化する能力
- LLM、VLMならびに最新のAI エコシステムに関する知識
歓迎スキル・経験(WANT)
- コンサルティング会社、システムインテグレータ、AIスタートアップ等での勤務経験
- 主要な国際会議(CVPR, ICML, NeurIPS等)での採択・発表実績
- 受託研究開発案件における顧客への技術提案・報告経験
- AI エージェントの構築・運用実績
[https://app.ashbyhq.com/api/images/user-content/1a65c548-6a4e-4805-84c0-451582ae3493/f3448c21-d1d0-493f-ac78-5e3d9c2781a9/re20230907_citadel_292a-2048x1365.jpg]