Overview
Associate, Digital Operations (LLM / AI Red-Team Evaluator) executing adversarial evaluation of large language model and AI systems as a specialist individual contributor on an AI red-teaming pilot.
What you'll do
- Act as a specialist individual contributor executing an AI red teaming adversarial evaluation pilot.
- Formulate risk hypotheses, design creative adversarial scenarios, and execute multi-turn, adaptive prompt attacks on authorized systems.
- Critically evaluate model outputs for grounding errors, source contradictions, stale information, and trust-boundary violations (such as indirect prompt injections).
- Capture, package, and document complete, reproducible evidence including exact prompt sequences, session/version context, and timestamps.
- Submit detailed finding cards for QA and calibrate judgments based on updated rubrics and guidelines.
- Separately classify candidate findings, quality issues, and inconclusive outcomes, knowing exactly when to stop and escalate unsafe or uncertain cases.
- Meet or exceed pre-production certification targets, including achieving a ≥85% agreement rate on adjudicated practice sets with zero material omissions.
- Maintain strong ownership of assigned attack surfaces and test surfaces within client-approved boundaries.
- Stay informed of client taxonomies, system changes, policy revisions, and evolving risk definitions.
- Collaborate with cross-functional teams including the Red-Team Team Lead, Project Manager, and QA analysts to resolve challenges constructively.
What you'll need
- 3+ years of experience.
- Graduate or undergraduate education.
- Core technical competencies: AI Red Teaming, Adversarial AI, Prompt Injection, Jailbreak Testing, Responsible AI, AI Governance, AI Safety.
- Core methodology competencies: LLM Evaluation, Model Evaluation, RLHF (Reinforcement Learning from Human Feedback), Human Feedback, Prompt Engineering, GenAI Testing, AI QA.
- Strong analytical, skeptical, and troubleshooting capability across LLM/RAG safety environments.
- Excellent written and verbal communication skills in US-English for structured reporting.
- High attention to detail and structured problem-solving approach to isolate false positives and verify citations.
- Demonstrated ability to build original, bounded attack hypotheses and execute multi-turn context escalation scenarios.
- Quality-focused mindset with accountability for complete evidence and defensible coverage over pure prompt volume.
Details
- Location: IND Hyderabad - Unit No. 201, 2nd Floor, Building 4 at Mindspace.
- Full-time employment.
Read the full description and apply on the company’s own careers page.