• Investigate and resolve complex challenges where no standard solution exists, specifically focusing on noise-robust NLP and data-efficient learning. • Design and manage methodologies for data generation, building pipelines to synthesize and process datasets for training and evaluation where real-world data is scarce or unlabeled. • Research and implement grounded agent systems, exploring techniques like ReAct, Chain-of-Thought, and tool-use optimization to reduce hallucination and improve planning reliability. • Design scientific evaluation protocols and benchmarks that go beyond standard metrics (like accuracy) to measure real world performance. • Stay current with the state-of-the-art in NLP and Agentic AI, contributing to internal knowledge sharing and external publications such as blogs or scientific articles.

GPTZero is on a mission to restore trust and transparency on the internet. As the leading AI detection platform, we empower educators, students, journalists, marketers, and writers to navigate the evolving landscape of AI-generated content. With millions of users and institutions relying on us, we’re building a category-defining company at the intersection of AI and information integrity. Our team comes from high-performing engineering cultures, including Meta, Perplexity, AWS, Affirm, and leading AI research labs, including Princeton, Caltech, and Vector Institute. GPTZero is launching a new product — AI Reviewer Customized Feedback We have over 10 million users scanning writing through GPTZero. In addition to AI and hallucination detection, we have a general writing feedback tool, but want to make the feedback experience more personal for users. We are building a library of editor templates. These will be feedback templates for a specific writing type such as ‘speechwriting.’ To build this template, we would run 5-10 speeches through the AI — and as an expert, you would provide edits to the general feedback to make it better for your specific case. Then, this will be featured in our library as a template credited and trained by yourself. What you’ll do Articulate how you evaluate writing quality in your domain Review and edit AI critique of ~5 writing samples using your professional criteria (ideally your own writing samples or writing samples you’ve edited if you are comfortable, but if not we can use public sources) Participate in a live workshop with a GPTZero researcher to apply and explain your edits Time commitment ~4 hours including 30 min live onboarding session with a member of the GPTZero team 3-3.5 hours of asynchronous template calibration Compensation $100 / hour (up to 4 hours per template) This research will directly inform how AI systems evaluate and critique writing across professional domains, with a goal to democratize personal editing rather than generic ChatGPT feedback. We are especially excited to collaborate and feature experts who care deeply about quality, standards, and craft. At GPTZero, our recruiting team is involved in every step of the hiring process. We use AI-based tools (such as Endorsed.ai and Juicebox.ai) to help us to accelerate candidates at the resume review stage by marking when candidates met certain key criteria. These tools are never the final say in a hiring decision - humans are.

Role Description A project dedicated to assessing and benchmarking advanced agentic audio models against leading systems. The program’s mission is to evaluate and optimize model performance for real-world customer support use cases. - Create and execute role-play–based evaluation scenarios that simulate realistic customer service interactions across multiple domains, including: - Flight bookings and travel support - Financial services - Telecommunications and technical support - Contribute to the development of diverse and representative datasets used to assess conversational audio agents. - Evaluate model performance across a standardized set of qualitative and quantitative metrics. - Ensure evaluations reflect real customer expectations for clarity, efficiency, and natural conversational flow. Qualifications - Strong fluency in English - Strong verbal communication skills in a simulated customer support context - Spanish proficiency including fluency across all language skills: reading, listening, writing, and speaking. Requirements - Basic computer programming literacy, including: - Understanding of JSON structures - Familiarity with functions and methods - Ability to reason about structured data and simple logic - Technical communication clarity when handling support-style problem-solving - Access to a high-quality microphone to ensure clean, reliable audio input during evaluations - Comfort working with structured prompts, evaluation rubrics, and technical guidelines - Device capable of running audio recording software and opening large technical documentation Benefits - We offer a pay range of $11-to- $30.65 per hour, with the exact rate determined after evaluating your experience, expertise, and geographic location. - Final offer amounts may vary from the pay range listed above. - As a contractor, you’ll supply a secure computer and high‑speed internet; company‑sponsored benefits such as health insurance and PTO do not apply. Company Description - Job title: Spanish Audio Specialist – AI Trainer - Employment type: Contract - Workplace type: Remote - Seniority level: Mid‑Senior Level

• Data Preparation - Prepare and maintain datasets for AI product evaluation, create annotation pipelines, build ground truth datasets, and support fine-tuning efforts with curated training examples • AI Model Evaluation - Test agent performance across model versions and prompt variations, build automated evaluation pipelines, and generate periodic reports that inform model selection and prompt engineering • Product Intelligence - Analyze user request patterns to identify high-demand use cases and capability gaps, cluster and classify large volumes of traces to find common themes and edge cases • User Analytics - Identify behaviors differentiating power users from casual users, analyze which user segments or industries drive highest engagement, and build statistical models to understand success vs failure drivers • Workflow Analytics - Optimize the first-time user funnel, correlate workflow usage patterns with customer retention, and identify power user behaviors across all workflows • Cost Optimization - Track quality metrics over time and alert on degradation, analyze token usage to identify cheaper model opportunities, and monitor latency across the agent pipeline (RAG, embeddings, generation) • Experimentation - Design and analyze statistically rigorous A/B tests for prompts and features, test new agent capabilities, and measure feature impact on user behavior and success rates