AI Guardrails Engineer

Mphasis; Concord (CA); On-site; USD 120,000 - 160,000; Full time

Job summary

Mphasis is seeking an AI Guardrails Engineer in Concord, CA, to design, build, and operate technical guards for AI systems. The role targets preventing harmful or non-compliant AI behavior with emphasis on LLM-powered features and agentic workflows in production environments.

You will address failure modes like prompt injection, data leakage, and unsafe tool usage, while aligning guardrails with product risk tiers and enterprise Responsible AI principles.

Responsibilities

• Design, build, and operate technical controls for AI systems.
• Prevent and detect harmful, insecure, non-compliant, or low-quality AI behavior.
• Focus on LLM-powered features, agentic workflows, and AI-assisted user experiences.
• Address new failure modes introduced by modern AI systems.
• Enable faster AI product delivery with lower risk and improved trust.
• Ensure compliance with internal policies and external regulations.

Job description

The AI Guardrails Engineer designs, builds, and operates technical controls(“guardrails”) that make AI systems safer, more reliable, policy-compliant, and predictable in production. This role focuses on preventing and detecting harmful, insecure, non-compliant, or low-quality AI behavior—especially in LLM-powered features, agentic workflows, and AI-assisted user experiences.

This role exists in software and IT organizations because modern AI systems introduce new failure modes (prompt injection, data leakage, hallucinations with high confidence, harmful or biased outputs, unsafe tool use, and policy violations) that cannot be solved by traditional application security or QA alone. The business value is enabling faster AI product delivery with lower risk, improved trust, reduced incidents, and demonstrable compliance with internal policies and external regulations.

Responsibilities

• Design, build, and operate technical controls(“guardrails”) for AI systems.
• Prevent and detect harmful, insecure, non-compliant, or low-quality AI behavior.
• Focus on LLM-powered features, agentic workflows, and AI-assisted user experiences.
• Address new failure modes introduced by modern AI systems.
• Enable faster AI product delivery with lower risk and improved trust.
• Ensure compliance with internal policies and external regulations.

Core Responsibilities

Strategic responsibilities

• Guardrails strategy and roadmap: Define a practical technical roadmap for AI guardrails aligned to product risk tiers, release plans, and enterprise Responsible AI principles.
• Risk-driven control design: Translate AI risk assessments into engineering requirements (prevent, detect, respond) across model, prompt, tool-use, and UI layers.
• Standard patterns and platforms: Establish reusable patterns (middleware, gateways, policy-as-code, eval harnesses) that reduce duplication across product teams.
• Safety-by-design in SDLC: Embed guardrails into design reviews, threat modeling, and release readiness criteria for AI features.

Operational responsibilities

• Production monitoring and alerting: Define and operate monitoring for unsafe content, policy violations, prompt injection attempts, sensitive data exposure, and model/tool misuse.
• Incident response for AI safety: Participate in on-call or escalation rotations (context-dependent) for AI safety/security incidents; lead technical mitigation and post-incident actions.
• Release governance support: Provide guardrails readiness checks, sign-offs, and evidence for staged rollout decisions (beta → GA).
• Continuous improvement loop: Use production signals, user feedback, and red-team outcomes to improve guardrails, prompts, filters, and detection models.