About this role
Rexzone seeks a remote, full-time AI prompt engineer to improve the reliability, groundedness and safety of LLM-based chatbots, copilots, agents and RAG systems. The role targets candidates based in India and lists annual pay of USD 63,360–126,720. The page’s country metadata separately says “US.”
Responsibilities
- Write, debug, version and maintain system prompts, developer prompts, reusable templates and prompt libraries. Translate product requirements into prompt patterns and integrate them with engineering teams.
- Run A/B tests and analyze outputs. Build evaluation rubrics and benchmark sets covering accuracy, completeness, instruction following, tone and policy compliance; monitor pass rates, inter-annotator agreement and regressions.
- Create preference and ranking data with rationales for RLHF-style and instruction-tuning workflows. Maintain annotation guidelines, including edge cases and escalation paths.
- Improve grounding and reduce hallucinations through RAG prompting, citations and traceability. Develop structured-output and tool-calling prompts, and safety tests addressing jailbreaks, policy violations and sensitive-data leakage.
Required qualifications
At least 3 years in software engineering, ML/NLP workflows or applied LLM prompt engineering; precise structured writing; experience with evaluation rubrics, pairwise ranking, golden sets and adversarial testing; familiarity with RLHF and human feedback; ability to interpret qualitative and quantitative results; and working knowledge of Python, notebooks, scripts, JSON and APIs.
Preferred qualifications
Experience with RAG pipelines, vector search, embedding evaluation and grounded-answer validation; content safety labels, policy taxonomies and guardrails; named entity recognition, information extraction or ontologies; LLM observability and evaluation tools; and collaboration with annotation vendors or BPO teams. Applicants are asked to submit a resume and a short portfolio of prompts, evaluation rubrics or experiment results.