[Tabelog] Design QA Autonomous Execution with AI Agents from Scratch | Create the Next Generation of Harnesses in a Large-Scale Environment | SET EngineerTranslated from Japanese

▼Job Description You will be responsible for designing and implementing harnesses that enable AI agents to autonomously execute QA tasks. 【Mission】 The Quality Assurance Department is a team that promotes the long-term vision "Continuous Quality by AI (CQ by AI)" aiming for "AI to consistently handle QA tasks." You will participate as an expert in providing QA knowledge and standards to the AI so that it can correctly judge quality. This is not just about streamlining testing, but an initiative to transform QA, which was a "pre-release checkpoint," into an autonomous mechanism embedded within the development process itself. We aim to achieve quality assurance where AI agents handle code reviews, risk analysis, and E2E tests for Tabelog, used by tens of millions of people monthly, with only final approval performed by humans. We have already achieved a code generation accuracy of 97%, a 52% reduction in test execution man-hours, and a 64% automation rate with automatic test coding, and we are now at the stage of building the next phase together. This position is responsible for preparing QA domain knowledge as guardrails, checklists, standards, etc., within the framework of harness engineering so that AI can make correct quality judgments. Unlike traditional QA engineers who "design and execute tests themselves and judge quality," this position shifts the role to "designing and verifying the criteria for AI judgment, and reviewing the AI's judgment results." While leveraging your expertise as a QA engineer, you will be responsible for the "job of teaching quality to AI," which is still rare in the industry. 【Specific Responsibilities】 - Practice Harness Engineering - Design and build AI agents for each QA process within the framework of harness engineering. - Risk Analysis / Test Analysis / Test Design Agents - Design and implement AI agents responsible for the upstream QA processes, analyzing specification documents and generating input for subsequent implementation and automatic test coding. - Automatic Test Coding and Automatic Test Execution Agents - Design and implement AI agents that form the core of QA tasks, implementing and executing test designs as automatic tests. - Design and implement CQ Feedback Pipeline - Accumulation and Analysis of Defect/Incident Information, and Construction of an Analysis Platform - Accumulate error logs, AI agent execution logs, defect reports, incident reports, etc., and build a platform for subsequent analysis and formulation of recurrence prevention measures. - Design and Implementation of Incident Analysis and Recurrence Prevention Agents - Design and implement AI agents that analyze incident information resulting from AI or human errors and implement them in a form of context or SKILL usable by AI agents. - Track and Disseminate External AI Trends - Track the latest trends in AI fields such as Harness Engineering and incorporate them into business design. - Disseminate internal practices and learnings externally through technical blogs, conferences, etc. 【Image of Joining the Company and Career Path】 ■ Immediately after joining (up to 3 months) - Understand Tabelog's services, business structure, and the overall architecture of CQ by AI. - Understand the mechanism of existing automatic test and automatic coding AI agents and address improvements. ■ 3 to 6 months after joining - Take charge of one or more AI agents of CQ by AI, design, and implement them. - Proactively promote the improvement of generated artifact accuracy within the autonomous execution cycle of AI agents. ■ 6 months to 1 year after joining - Aim for the state where the assigned AI agent can execute fully autonomously. - Reach a state where practical knowledge can be disseminated through technical blogs and internal cross-functional expansion. ■ Career Path - QA AI Architect: A senior specialized position leading the harness design and evaluation platform for CQ by AI. - AI Quality Foundation Lead: A lead position overseeing cross-functional expansion to multiple pipelines and multiple domains. - There is also a career path to establish a presence in AI quality engineering within the industry through external dissemination of practices and learnings both internally and externally. AI-specialized QA roles have higher market value compared to traditional QA, and choosing this career path directly leads to an increase in market value. ▼Required Skills/Experience - Experience in designing/implementing automated tests (e.g., Playwright / Selenium / Cucumber) and integrating/maintaining them in CI/CD pipelines (2+ years). - Practical experience in backend development (API design/implementation, RDBMS design/operation, etc.) or SRE (cloud infrastructure such as AWS / GCP, IaC, CI/CD operation, etc.) (2+ years). - Experience in designing/implementing autonomous execution pipelines connecting multiple AI agents, and implementing AI agent and tool integration using MCP Server. - Experience using Claude Code (or AI coding agents like Cursor) as a daily development environment for over 6 months. - Experience in analyzing business operations, designing/formalizing the scope for AI to handle, and creating mechanisms for autonomous execution without human intervention. ▼Welcome Skills/Experience - Experience in upstream QA tasks such as test strategy formulation and quality standard design, or deep operational knowledge in QA/SRE domains (incident management, failure response, etc.). - Experience using LLM orchestration frameworks such as LangChain / LangGraph, or ML frameworks such as PyTorch / TensorFlow. - Experience in log and data analysis using SQL / BigQuery. - Track record of external dissemination through technical blogs, conferences, etc. ▼Desired Candidate Profile - Harness Thinking: Naturally considers implementing recurrence prevention in "environmental design" rather than "human attention." - Business Understanding First: Prioritizes understanding the "actual state of business" and the "structure of problems to be solved by AI" before focusing on technology. - POC Oriented: Quickly builds working prototypes and makes decisions based on empirical evidence rather than lengthy discussions. - Collaboration with Domain Experts: Able to explain designs to non-engineers and build standards together. - External Communication Awareness: Values external dissemination of internal practices and learnings, and cherishes dialogue with the industry. ▼Job Appeal 【Create a World-First Initiative from Scratch】 - The system where "QA tasks are autonomously executed by AI agents" is not yet established in the industry. You can implement the concept of Harness Engineering in practical work and create a precedent yourself. 【Demonstrate on a Large-Scale Service】 - You can build a system where AI handles incident management, test generation, and code reviews for Tabelog, used by tens of millions of people monthly. We have already achieved a code generation accuracy of 97%, a 52% reduction in test execution man-hours, and a 64% automation rate with our automatic test coding AI agent. You can build the next stage in an environment with proven results. By demonstrating in a scaled environment, you can build core achievements that pave the way for the next stage. 【Get Ahead of the Shift from "Code Writer" to "Environment Designer"】 - Models are processing power, harnesses are the OS. You are in a position to define your own work at the forefront of the changing role of engineers. 【Leverage Both Engineering and QA Domain Expertise】 - This job can only be accomplished by combining backend and infrastructure technical skills with QA operational knowledge. You will be in a position to solve problems that cannot be solved by either skill alone, as an engineer possessing both. ▼Development Environment ■ AI Agents: Claude Code (SKILL / Sub-agent / Hooks / Agent Teams) / Devin / MCP Server ■ Test Automation: Selenium, Appium, Cucumber, Playwright ■ CI/CD: CircleCI, Bitrise ■ Data Platform: BigQuery, Tableau ■ Task Management & Analysis: Asana ■ Communication: Slack, Teams, GitHub, Confluence, Miro ■ Development Tech Stack: Ruby on Rails, React.js, Next.js, Swift/SwiftUI, Kotlin ▼Employment Type Full-time employee (No fixed term) ▼Probationary Period 3 months * No change in employment status or treatment during the probationary period. ▼Salary Annual salary: ¥8,000,000 - ¥12,000,000 ▼Bonus Bonus once a year (June) ▼Annual Salary Lower Limit (JPY) 8,000,000 ▼Annual Salary Upper Limit (JPY) 12,000,000 ▼Work Location Head Office: 3-5-7 Ebisu-minami, Shibuya-ku, Tokyo, Digital Gate Building * We employ a hybrid work style that flexibly combines remote work and office attendance depending on job content and team circumstances. ▼Working Hours Flextime system (standard working hours: 8 hours per day) - Core time: 10:00 - 15:00 * Overtime may occur due to work circumstances. * This differs for management supervisors and those under discretionary labor systems. * You can flexibly change your start and end times according to team circumstances and your own work pace. ▼Break Time 1 hour during working hours ▼Overtime Work Yes ▼Holidays, Paid Leave, Special Leave Complete two-day weekend system - Scheduled holidays: Saturdays, Sundays, National Holidays, Year-end and New Year holidays (12/29 - 1/3) - Leave: Summer vacation (3 days), Paid leave, Condolence leave * Paid leave: Up to 10 days granted depending on the month of joining. ▼Benefits We aim to be a company where employees can experience both the "fun of working" and the "ease of working," which Kakaku.com has valued since its founding. We will continue to actively review our systems, reflecting employee feedback while considering usage. - Social insurance (employment, health, worker's accident, employee pension) - Defined contribution pension plan - Group life insurance - Free comprehensive medical check-ups (regular health examinations) - EAP counseling program - Congratulations and condolences money - Maternity and childcare leave (can be taken until the end of the fiscal year when the child turns 3 years old) - Childcare short-time work system (can be taken for up to 12 years until the child graduates from elementary school; option to choose a coreless flextime system) - Child care leave (10 days per year, 5 of which are paid leave; 20 days per year for multiple children, 10 of which are paid leave) - Nursing care leave - Volunteer leave - Family allowance (conditional on payment) - Remote work environment allowance ▼Allowances - Commuting expenses fully reimbursed (with limits) - Family allowance (conditional on payment) - Remote work environment allowance ▼Insurance Enrollment - Social insurance (employment, health, worker's accident, employee pension) - Group life insurance ▼Measures to Prevent Passive Smoking at the Workplace Passive smoking measures for employees: Yes Measures: Indoor smoking prohibited in principle (smoking area available) ▼Education/Skill Improvement/Career Development - Self-study support system available [Support for Engineers] - Study sessions: Company-wide study sessions, including those for Kakaku.com, are held once every three months. Study sessions for all Tabelog engineers are held on average once a month. Study sessions within each engineer team at Tabelog are also held irregularly. - Purchase of technical books: While they become company assets, technical books can be purchased freely within the budget. Long-term loans are also possible. Many engineers purchase technical books every month. - Seminar participation: Participation is possible during working hours. If participating in paid seminars, the company will cover the costs after prior consultation. - Support for organizing internal and external events: We provide support such as venue provision and sponsorship. Tabelog sponsors RubyKaigi, held annually. ▼Health - Free comprehensive medical check-ups (regular health examinations) - EAP counseling program ▼Interview/Selection Process Document screening -> 1st interview -> Final interview -> Offer * The selection flow is subject to change.