Senior Principal AI Systems Engineer

Advantest is seeking a highly experienced Senior Principal AI Systems Engineer to architect and deliver
advanced production AI capabilities. This role is for a hands-on technical leader with deep expertise in
agentic AI, open-source model development, model fine-tuning, evaluation systems, and reliable AI
infrastructure.
The ideal candidate combines advanced AI knowledge with strong software-engineering discipline and
has experience building systems that reason across multiple steps, use tools, evaluate results, recover from
failures, and improve from feedback.
Key Responsibilities
Agentic AI Systems
* Architect production-grade systems for tool-using and multi-step AI agents.
* Design orchestration, planning, memory, session management, retries, timeouts, observability, and failure
recovery.
* Establish structured and typed interfaces between models, tools, data services, and applications.
* Build controls that prevent agents from bypassing required validation and approval stages.
* Develop reusable agent frameworks that support multiple products and deployment environments.
Model Development and Fine-Tuning
* Fine-tune open-source language models and specialized models for complex technical applications.
* Apply LoRA, QLoRA, full fine-tuning, distillation, preference optimization, and related post-training
methods.
* Develop efficient strategies for adapting foundation models to new applications and datasets.
* Evaluate tradeoffs among model quality, inference performance, deployment cost, security, and
maintainability.
* Build optimized models and inference profiles for constrained deployment environments.
Evaluation and Quality
* Define measurable standards for model accuracy, reliability, safety, and production readiness.
* Build offline evaluation datasets, automated regression suites, judge systems, and promotion gates.
* Develop methods for measuring confidence, consistency, tool-use accuracy, and action quality.
* Establish processes for model comparison, controlled release, rollback, and continuous improvement.
* Ensure model outputs remain grounded in available evidence and approved data sources.
Training Data and Learning Pipelines
* Design reproducible pipelines for training-data generation, cleaning, labeling, versioning, and validation.
* Develop synthetic-data and preference-data strategies where appropriate.
* Implement controls for data leakage, contamination, duplication, provenance, and customer isolation.
* Convert expert feedback and observed outcomes into high-quality training and evaluation datasets.
* Maintain traceability between datasets, experiments, model versions, and production results.
AI Platform and MLOps
* Build repeatable training, evaluation, and deployment workflows for multi-GPU infrastructure.
* Establish model lifecycle practices, including model cards, release criteria, monitoring, and rollback.
* Support secure, private, air-gapped, and customer-controlled deployment environments.
* Develop observability for model behavior, tool execution, latency, cost, and failure conditions.
* Partner with platform and CI/CD teams to make AI workflows repeatable, testable, and auditable.
Technical Leadership
* Set technical direction for a small, highly skilled AI engineering team.
* Review architectures, models, training methods, and production implementation decisions.
* Mentor engineers and establish durable AI engineering practices.
* Work with domain experts to translate complex technical requirements into reliable AI capabilities.
* Communicate technical risks, tradeoffs, progress, and recommendations to engineering and executive
stakeholders.