AI Research Engineer

Summary

Join Tether's AI model team and drive innovation across the entire AI lifecycle. Develop and implement rigorous evaluation frameworks and benchmark methodologies for pre-training, post-training, and inference. Design metrics and assessment strategies to ensure models are highly responsive, efficient, and reliable. Work on various systems, from resource-efficient models to complex, multi-modal architectures. Collaborate with cross-functional teams to share evaluation findings and integrate stakeholder feedback. Engineer robust evaluation pipelines and performance dashboards. Set industry-leading standards for AI model quality and reliability, delivering scalable performance and tangible value.

Requirements

• A degree in Computer Science or related field
• Ideally PhD in NLP, Machine Learning, or a related field, complemented by a solid track record in AI R&D (with good publications in A* conferences)
• Demonstrated experience in designing and evaluating AI models at multiple stages from pre-training, post-training, and inference
• You should be proficient in developing evaluation frameworks that rigorously assess accuracy, convergence, loss improvements, and overall model robustness, ensuring each stage of the AI lifecycle delivers measurable real-world value
• Strong programming skills and hands-on expertise in evaluation benchmarks and frameworks are essential
• Familiarity with building, automating, and scaling complex evaluation and benchmarking pipelines, and experience with performance metrics: latency, throughput, and memory footprint
• Proven ability to conduct iterative experiments and empirical research that drive the continuous refinement of evaluation methodologies
• You should be adept at staying abreast of emerging trends and techniques, leveraging insights to enhance benchmarking practices and model reliability
• Demonstrated experience collaborating with diverse teams such as product, engineering, and operations in order to align evaluation strategies with organizational goals
• You must be skilled at translating technical findings into actionable insights for stakeholders and driving process improvements across the model development lifecycle

Responsibilities

• Develop, test, and deploy integrated frameworks that rigorously assess models during pre-training, post-training, and inference
• Define and track key performance indicators such as accuracy, loss metrics, latency, throughput, and memory footprint across diverse deployment scenarios
• Curate high-quality evaluation datasets and design standardized benchmarks to reliably measure model quality and robustness
• Ensure that these benchmarks accurately reflect improvements achieved through both pre-training and post-training processes, and drive consistency in evaluation practices
• Engage with product management, engineering, data science, and operations teams to align evaluation metrics with business objectives
• Present evaluation findings, actionable insights, and recommendations through comprehensive dashboards and reports that support decision-making across functions
• Systematically analyze evaluation data to identify and resolve bottlenecks across the model lifecycle
• Propose and implement optimizations that enhance model performance, scalability, and resource utilization on resource-constrained platforms, ensuring efficient pre-training, post-training, and inference
• Conduct iterative experiments and empirical research to refine evaluation methodologies, staying abreast of emerging techniques and trends
• Leverage insights to continuously enhance benchmarking practices and improve overall model reliability, ensuring that all stages of the model lifecycle deliver measurable value in real-world applications