We are looking for a cybersecurity specialist to help safely implement and support such solutions.
The role is at the intersection of Application Security, DevSecOps, Cloud/Kubernetes Security, and GenAI/LLM/AI agent security.
What you will do:
- Conduct security reviews of AI/LLM solutions: LLM applications, AI agents, RAG, vector databases, integrations with external APIs and cloud LLMs;
- Participate in threat modeling of AI solutions: prompt injection, jailbreak, data leakage, insecure tool calls, excessive AI agent privileges, RAG/data poisoning;
- Check for the presence and correctness of protective mechanisms: guardrails, personal data filters, prompt/response length limits, rate limiting, response validation, tool call control;
- Evaluate data security: personal data, confidential information, data sources for RAG, embeddings, vector database indexes;
- Verify that AI agents operate on the principle of least privilege and do not perform sensitive operations without manual confirmation;
- Check the isolation of user sessions, contexts, and memory of AI solutions;
- Participate in checking the security of local LLM deployment in Kubernetes: containers, RBAC, secrets, network policies, ingress, resource limits;
- Formulate requirements for logging, auditing, and tracing of AI solutions: trace ID, prompt/response logs, model calls, tool calls, AI agent actions;
- Develop test scenarios for checking AI solutions for prompt injection, data leakage, bypassing limitations, and insecure agent behavior;
- Assist development and operations teams in safely using local and cloud LLMs.
Requirements:
- Cybersecurity experience of 2–3 years in one or more areas: Application Security, DevSecOps, Cloud Security, Kubernetes Security, API Security, Product Security;
- Understanding of security review, threat modeling, and secure development principles;
- Understanding of web/API security: authentication, authorization, OAuth2/OIDC, SAML, JWT, RBAC, service-to-service interaction;
- Basic understanding of the AI/LLM stack: LLM, prompt, system prompt, context window, embeddings, RAG, vector database, AI agent, agent memory, tools/function calling;
- Understanding of key risks in LLM applications: prompt injection, indirect prompt injection, jailbreak, sensitive data leakage, system prompt disclosure, insecure model response handling, excessive agency;
- Understanding of security specifics for RAG and vector databases: e.g., Milvus, Qdrant, Weaviate, pgvector, Chroma, Elasticsearch/OpenSearch vector search;
- Understanding of risks associated with personal and confidential data when using local and cloud LLMs;
- Basic understanding of Kubernetes and container security: secrets, service accounts, RBAC, network policies, resource limits;
- Ability to formulate clear and verifiable security requirements for product teams.
What we are looking for in candidates:
- Experience in security review of AI/LLM, RAG, chatbots, AI assistants, or agent solutions;
- Familiarity with OWASP Top 10 for LLM Applications, OWASP Agentic AI Threats, MITRE ATLAS, NIST AI RMF;
- Understanding of security for agent tools/function calling: allowlist tools, input parameter schemas, privilege control, human-in-the-loop;
- Experience with SIEM/security logging: designing security events, auditing actions, incident investigation;
- Understanding of approaches to secure execution of model-generated code: sandbox, container isolation, restriction of access to secrets, network and resource limits;
- Basic Python skills, working with JSON/YAML, API testing, or security check automation.
Will be a plus:
- Experience with OpenAI/Azure OpenAI, AWS Bedrock, Google Vertex AI, Anthropic, or other LLM providers;
- Experience with local open-source LLMs and inference services;
- Experience setting up or checking guardrails, personal data filters, DLP approaches, content moderation;
- Experience participating in AI red teaming or developing tests for prompt injection and data leakage.
What SberZdorovye can offer you:
- Format: Hybrid (from 2 days a week, Technopark metro station).
- Health Care: Free access to telemedicine and in-person appointments at private clinics (including dentistry and psychology sessions);
- Benefits: Mortgage loan conditions from Sberbank.
- Development: Individual development plan for everyone, which will help you always see the direction of your path;
- Training: English language with a corporate tutor and discounts on Skyeng courses, compensation for training from leading market platforms;
- Stability and Comfort: Official employment, stable payments, sick leave and vacation pay, modern equipment;
- Sports: Partial compensation for sports club membership, free squash and football sessions.
