AI Evaluation Engineer
Employment Information
Report this job
Job expired or something wrong with this job?
Job Description
AI Evaluation Engineer building statistical evaluation, CI/CD testing, and observability infrastructure for Scopic’s production AI agents. Designing adversarial, multi-turn, and structured-output test systems.
Responsibilities:
- Design and deploy automated CI/CD testing pipelines that validate agent behavior before deployment
- Build and maintain golden datasets of regression prompts, edge cases, and synthetic scenarios to measure agent accuracy, safety, and latency
- Implement fast, low-cost code-based checks, including JSON schema validation, regex matching, API status checks, and embedding similarity
- Design test suites for multi-turn conversations, validating context retention, state drift, and safeguards against infinite tool-calling loops
- Create adversarial test suites covering prompt injections, malformed inputs, and chaotic tool responses to secure sandboxes and guardrails
- Implement tracing with OpenTelemetry, LangSmith, Phoenix, or similar tools
- Reduce LLM API costs through caching and mocking
Requirements:
- Proficiency in Python or TypeScript, with a focus on API testing, headless execution, and asynchronous systems
- Hands-on experience building LLM evaluation pipelines or statistical testing frameworks, such as confidence intervals or pass rates over N runs
- Experience integrating tests into GitHub Actions or similar CI systems using SDK-level execution
- Expertise in validating structured outputs: JSON schema, OpenAPI specs, and function-calling payloads
- Familiarity with tracing tools such as OpenTelemetry, LangSmith, Arize Phoenix, W&B Weave, or custom logging
- Ability to independently design and own an evaluation architecture from scratch
- Strong written and spoken English
- At least 3 hours of daily overlap with US Central Time business hours
- Experience with code execution sandboxes or agent frameworks such as LangChain, LlamaIndex, AutoGen, or CrewAI
- Deep understanding of embedding models and vector databases
- Experience load testing non-deterministic endpoints
Benefits:
- Paid training and other professional opportunities
- Flexible schedule and preferred working hours
- Ability to work from anywhere with a stable internet connection
- Consistent flow of engaging, challenging product work
- Annual pay increases for diligence and strong performance
- Monthly compensation payment via wire transfer
Similar Jobs
English AI Safety Specialist
AI Safety Specialist reviewing English conversations and labeling AI responses. Helping Mercor’s AI lab and enterprise partners train and enhance frontier models.
Clinical Evaluator, AI Products – Contract
Clinical Evaluator testing Headway’s AI mental-health products for clinical accuracy and safety. Refining evaluation rubrics and guiding product launch decisions.
AI Harness Engineer
AI Harness Engineer building reliable AI tools, agent harnesses, and P2P systems for Tether’s local AI assistants. Advancing QVAC within Tether’s blockchain-powered digital finance ecosystem.
AI Harness Engineer
AI Harness Engineer building agent harnesses and P2P technology for Tether’s local AI products. Delivering reliable tools, modules, and scalable integrations across model types.
AI Safety Red Teamer
AI Safety Red Teamer adversarially testing frontier AI models for Mercor, which partners with AI labs to train and enhance frontier systems. Identifying vulnerabilities across high-risk domains.
AI Prompt & Workflow Specialist
AI Prompt & Workflow Specialist designing and optimizing generative AI solutions for ExamWorks’ healthcare services operations. Translating business requirements into tested prompts, agents, and automation workflows.
Quality Control Specialist – AI/ML, Ukrainian Language
Quality control specialist reviewing Ukrainian AI speech transcripts for TSMG. Correcting display-text formatting, punctuation, capitalization, and normalization issues.
AI Enablement Specialist
AI Enablement Specialist evaluating and implementing AI tools, automated workflows, and process improvements for Wilson Sonsini’s legal practices. Leading pilots, adoption, training, and cross-functional innovation initiatives.
Head of Teaching – AI & Future Skills
Head of Teaching leading nationwide remote education operations for an EdTech scale-up. Shaping AI-focused curricula, teaching quality, processes, and scalable learning programs.
Clinical Review RN, Oncology AI Review
Clinical Review RN validating AI-driven oncology workflows for Natera, a global cfDNA testing company. Reviewing complex cases, payer policies, and AI classifications.
AI Safety Red Teamer
AI Safety Red Teamer stress-testing frontier AI models for Mercor, which partners with AI labs to train and enhance AI systems. Uncovering jailbreaks, hallucinations, and policy failures.
AI Safety Red Teamer
AI Safety Red Teamer stress-testing frontier AI models through adversarial prompts. Identifying vulnerabilities and improving model alignment, robustness, and safety for Mercor’s AI lab partners.
AI Safety Red Teamer
AI Safety Red Teamer identifying vulnerabilities in frontier AI systems for Mercor's partner AI labs. Designing adversarial prompts and evaluating high-risk model behavior.
Scientific Language & AI Adaptation Contributor – Korean, Japanese, Arabic
Scientific language contributor helping Maxibook, an AI education startup, adapt English scientific content across Korean, Japanese, and Arabic. Validating AI outputs for conceptual accuracy, naturalness, and academic credibility.
AI Safety Red Teamer
AI Safety Red Teamer stress-testing frontier AI systems for Mercor. Designing adversarial prompts and documenting vulnerabilities to improve model safety and alignment.
Research Staff, Voice AI Foundations
Research Staff pioneering latent-space models, neural audio codecs, and speech systems at Deepgram. Advancing scalable, efficient voice AI for real-time human-machine interaction.
Research Staff – Voice AI Foundations
Research Staff developing latent-space models for Deepgram’s voice AI platform. Advancing neural audio codecs, generative speech, multimodal systems, and hardware-efficient inference.
Research Staff, Voice AI Foundations
Research Staff pioneering latent-space models for codecs, speech generation, and multimodal voice AI. Advancing Deepgram’s real-time speech platform through scalable, efficient foundational research.
AI Operations Engineer
AI Operations Engineer building and supporting enterprise AI agents, chatbots, and integrations. Enabling secure, governed AI operations for Turtle’s electrical distribution business.
AI Evaluations Engineer – School of Government and Policy
Remote AI evaluations engineer managing sensitive data, audits, queries, and reports. Supporting Johns Hopkins’ School of Government and Policy with data processes and user training.