Software Engineer

Objectives

PMAX is building an AI agent capability across the company. Teams are writing decision playbooks, testing them through structured feedback loops, and turning them into repeatable, improvable tools. The first wave of agents is being built now using chatbots and spreadsheets. Your job is to build the infrastructure that takes these proven agents from manual copy-paste to reliable, deployed systems. You will be the first dedicated AI engineering hire. For the first 6–12 months, you are a team of one working closely with the CTO and domain builders across the company. You will not be building agents yourself—domain experts own the logic and the playbooks. You build the rails they run on: the APIs, the UIs, the auth, the logging, the deployment pipeline. When a team has a proven agent that needs to scale beyond copy-paste, you make that happen. This is a building role, not a research role. You will ship working systems, not write papers or train models.

Job Description

• Take proven agents to production. When a team has built and tested an AI agent manually and it’s ready for wider use, you build the system around it: API layer, simple UI, auth, deployment. You assess whether existing prototypes can be hardened or should be rebuilt from the documented playbook and decision rules.
• Build workflow automation. Wire LLM APIs to data sources (Google Sheets API, REST endpoints). Build orchestration: scheduled runs, trigger-based execution, output routing. Add proper error handling, retry logic, and fallback behaviour so agents run reliably without manual intervention.
• Set up and own the infrastructure layer. Deployment pipeline (Vercel or similar). Auth layer (Cloudflare Access with company SSO). Logging that captures inputs, outputs, user actions, and override reasons. Environment variable management. Reusable project templates.
• Instrument everything. Every agent run must be auditable: what went in, what came out, which version of the decision framework was used, what the user did with the output. This logging is the company’s primary learning signal. Build it well.
• Build protected interfaces for sensitive agents. Some agents contain commercially sensitive decision logic (pricing, margins, competitive positioning). Build interfaces where users can interact with the agent and get results without seeing the underlying framework.
• Own security standards for AI tooling. Review deployments built by non-engineers across the company. Ensure keys aren’t hardcoded, SSO is configured, data access is controlled. This is not glamorous. It is essential.
• Support and mentor builders. As more people across the company build AI tools (including with AI-assisted coding), you become the person who reviews their work, helps with architecture decisions, and ensures quality and security standards are met.

Job Requirements

• LLM API integration (non-negotiable). You have built and shipped at least one application that calls an LLM API (OpenAI, Anthropic, Google) in a real-use context—not a hackathon demo. You understand prompt management, token economics, streaming, error handling, and retry logic. You know what a system prompt is and why it matters.
• Full-stack web development. TypeScript/Node.js or Python, ideally both. You can build a functional UI (Next.js or similar), wire up an API layer, and deploy it. You don’t need to be a frontend specialist—clean and functional beats pretty.
• Deployment and auth. You have deployed applications to Vercel, Railway, Render, or similar. You can configure SSO (Cloudflare Access or equivalent). You manage secrets properly (.env, platform environment variables, not hardcoded).
• Data integration. You can read from and write to Google Sheets API, REST APIs, and simple databases (Postgres, SQLite). You understand data contracts: schema definitions, column naming conventions, source-of-truth discipline.
• 3–5 years of professional software development experience. You have shipped and maintained software in a team. You understand version control, code review, and why logging matters. You have been on-call or owned a system in production.

Strongly Preferred

• Experience building internal tools or B2B products. You’ve been close to the business problem, not just executing tickets. You understand that the domain expert knows things you don’t, and you build for them, not around them.
• Experience working as a solo engineer or in a very small team. You are comforta…