Job Details

AI Security Engineer at T-Bank. Moscow. Salary is negotiable. Development of traffic inspection logic between applications and LLM, research into AI agent behavior, testing and validation of controls.

• Development of traffic inspection logic between applications and LLM: protocol parsing, content decryption, and analysis; • Research into AI agent behavior (tool calling, MCP, multi-step workflows) and creation of detections for anomalous and malicious patterns. Testing and validation of controls; • Development of autotests for security guards: verifying that filters, detections, and policies work correctly and do not degrade between releases; • Negative testing: checking for false positives, resistance to bypasses, regression during model and rule updates; • Formation and maintenance of datasets for testing detections (adversarial prompts, edge cases, evasion techniques); • Participation in the development of the AI traffic processing and classification pipeline.

• Understanding of the network stack: HTTP/HTTPS, TLS termination, ICAP, content inspection; • Ability to understand others' code and undocumented protocols — a significant part of the work involves research and reverse engineering; • Experience in security testing: fuzzing, property-based testing, writing security regressions; • Understanding of how LLM inference works and practical experience with agentic frameworks (LangChain, CrewAI, AutoGen, Claude Code, OpenAI Agents SDK, etc.) will be a plus.