C

Cisco

Makes networking gear and security software

AI Safety & Reasoning Engineer - Reasoning Transparency

AI ResearchPosted on Nov 13
Full-timeHybrid

About the Job

Requirements

  • A Bachelor's degree and 7 years of related experience, a Master's degree and 4 years of related experience, a PhD and 1 year of related experience, or equivalent related work experience is required.
  • Hands-on experience deploying or operating machine learning models in a production or production-like environment using Python is required.
  • The candidate must be able to conduct rigorous empirical work and communicate technical results, assumptions, and limitations to cross-functional collaborators.
  • A practical understanding of transformer-based language models, inference-time computation, and alignment approaches such as reinforcement learning from human feedback or reinforcement learning from AI feedback is required.
  • Proficiency in building evaluation frameworks that use adversarial prompt engineering to probe model limits and ensure model reliability and safety is required.

Responsibilities

  • Operationalize transparency by translating concepts from mechanistic interpretability and eliciting knowledge into scalable pipelines that run alongside model inference.
  • Build infrastructure to extract, map, and interpret model activations and latent representations, supporting analysis of how internal signals relate to model outputs.
  • Design red-teaming frameworks and automated behavioral-auditing stress tests to identify misalignment, sycophancy, deceptive behavior, and coordination risks across models or agents.
  • Partner with machine-learning engineering to integrate transparency probes while managing inference latency, memory use, reliability, and operational cost.
  • Track relevant AI safety research to rapidly prototype, evaluate, and deploy promising transparency methods.
  • Define evaluation methods to communicate evidence, limitations, and actionable findings to research, engineering, and product partners.

Desired Qualifications

  • High proficiency in Python frameworks, including experience building performant code for distributed training or high-throughput inference environments.
  • Hands-on experience probing, mapping, or analyzing model activations, circuits, representations, or other internal states.
  • Experience deploying and operating machine-learning pipelines with container orchestration such as Kubernetes, continuous integration and continuous delivery tools, and model tracking tools.
  • Experience applying information security principles or structured threat modeling to AI systems, model infrastructure, or sensitive data.
  • Familiarity with methods for eliciting model knowledge or beliefs and evaluating cases where stated outputs may diverge from internal signals.

About the Company

Cisco

Makes networking gear and security software

Cisco makes the switches and routers that connect offices, data centers, and cloud services, and it also sells cybersecurity tools and Webex video meetings to businesses and governments. It was one of the first companies to sell routers that let different kinds of computer networks exchange data, and in 2000 it briefly became the world's most valuable company. Customers run its network gear, from Wi-Fi access points to AI data center systems, with Cisco software that sets up and secures the network, while its Splunk unit searches computer logs to catch outages and attacks. Cisco earns money from hardware sales, software subscriptions, and support contracts. It was founded in 1984 and is based in San Jose, California.

Company size
10,001+
Company stage
IPO
Headquarters
San Jose, CA
Founded
1984

Recently Posted Jobs

Explore more opportunities at Cisco

See more jobs