C

CrowdStrike

Builds cloud software that stops cyberattacks

AI Research Scientist - Agentic Systems

AI Research +1Confirmed live in the last 24 hours
Full-timeRemote in USARemote

About the Job

Requirements

  • Strong foundations in machine learning, probability, and statistics, including understanding uncertainty, statistical skew and variance, and experimental design.
  • PhD-level depth in modern machine learning research or equivalent mastery, including the ability to read, critique, implement, and improve upon current papers; a doctorate is not required.
  • Experience training generative models and strong command of large language model training fundamentals, including architecture, optimization, tokenization, data, and scaling behavior.
  • Reinforcement learning and post-training skills, including RLHF/RLAIF, PPO/GRPO/DPO, reward modeling, and building reinforcement learning environments for agents.
  • Experience building agentic systems, including ReAct, planning, reflection, tool and function calling, and retrieval, memory, and context management.
  • Experience with systematic prompt optimization and designing and building evaluations for large language model systems.
  • Fluency with GPUs, PyTorch, and large language model training and serving tools such as Hugging Face Transformers, TRL, PEFT, DeepSpeed, FSDP, vLLM, TGI, and SGLang.
  • Strong, reproducible research engineering skills, including clean Python and disciplined experiment tracking.
  • Ability to work independently on ambiguous and complex objectives and communicate clearly within a large project team.
  • Proven experience using artificial intelligence technologies to enhance decision-making, streamline workflows and processes, improve efficiency, and drive business outcomes.

Responsibilities

  • Work at the intersection of artificial intelligence and threat research.
  • Work with cybersecurity subject-matter experts to understand analyst workflows and security operations procedures.
  • Post-train large language models and agents using supervised fine-tuning and reinforcement learning methods, including RLHF/RLAIF, PPO/GRPO/DPO, and reward modeling, to automate analyst procedures and improve reliability on real security tasks.
  • Develop AI agents and combine them into complex workflows involving planning and reasoning loops, tool and function calling, retrieval, and memory.
  • Research approaches to agentic planning and prototype state-of-the-art methods from the literature.
  • Establish objective criteria for benchmarking agentic systems, including evaluations, LLM-as-judge pipelines, and trajectory-level metrics, using statistical rigor.
  • Optimize prompts and inference to improve model performance.
  • Collaborate and coordinate across Engineering, Data Science, and Managed Services teams, and partner with engineers to move prototypes toward production.
  • Track developments in artificial intelligence and help identify, define, and prioritize research areas.

Desired Qualifications

  • Experience generating training data and environments, including synthetic data, agent trajectories or rollouts, and task simulators.
  • Familiarity with inference-time scaling or test-time compute, including search, self-consistency, verifier-guided decoding, and long chain-of-thought.
  • Experience with agent safety and guardrails, including sandboxing, abuse and jailbreak resistance, and reliability for autonomous systems.
  • Skill in interpretability and failure analysis, including diagnosing why a model or agent fails.
  • Notable open-source contributions and technical writing ability.
  • Passion for cybersecurity and understanding of the problem space, or interest in applying machine-learning skills to cybersecurity; a security background is a plus but not required.

About the Company

CrowdStrike

Builds cloud software that stops cyberattacks

CrowdStrike builds cybersecurity software for organizations that need to protect computers, cloud workloads, identities and data from digital attacks. Its Falcon platform combines these defenses in one cloud system, using a lightweight sensor on devices to collect activity and detect suspicious behavior. Security teams investigate alerts and automate responses through a shared console, while the platform connects information from different parts of a business to reveal attacks that isolated tools miss. Customers pay subscription fees based on the products and assets covered, and CrowdStrike also charges for incident response services. CrowdStrike was founded in 2011 and is based in Austin, Texas.

Company size
10,001+
Company stage
IPO
Headquarters
Austin, TX
Founded
2011

Recently Posted Jobs

Explore more opportunities at CrowdStrike

See more jobs