AI Developer

Level: Software Engineer II / Senior Software Engineer Position Summary: Our client is building the AI-native engineering organization of the future — one where small, durable teams direct AI to deliver software faster and more reliably than traditional models allow. This role sits inside that effort from day one: working with the latest in agentic AI development, helping shape the practices and platforms the next generation of technology delivery will be built on, and contributing directly to a capability that compounds in value with every project completed. AI Developers are engineers who drive agentic development day to day. They direct Claude Code through the plan-first workflow, author the plans it executes, review and correct its output, and remain fully accountable for what ships. Their leverage comes from operating as directors of a capable engineering partner rather than as line-by-line authors — while retaining the engineering depth to know when the AI is wrong. A standard pod carries one to two AI Developers. Success is measured not by lines of code written but by the reliability and quality of working software reaching production. Responsibilities:

AI-DIRECTED DEVELOPMENT

Run the plan-first workflow — on every non-trivial change: explore, plan, review, then implement — never jumping straight to code. Direct Claude Code effectively — and review its output as rigorously as a senior engineer would review a teammate’s pull request. Catch and correct AI errors — wrong problem solved, unintended side effects, standards drift — before they reach review. Understand, contribute and Improve AI-DLC workflow and skills — created inside plan mode and review before any code is written.

ENGINEERING STANDARDS & QUALITY

Keep implementation on-standard — no raw SQL, standard response envelope, kebab‑case URLs, OTEL instrumentation, no writes to the legacy boundary. Maintain CLAUDE.md and settings.json — the pod’s AI context file and permission boundaries; update as conventions evolve. Practice context-window discipline — one task per session, dump to PLAN.md and /clear near the 60% threshold, commit at least hourly. Keep pull requests under ~400 lines — and scope them to one feature or one slice.

CONTINUOUS IMPROVEMENT

Use the company standards, slash commands /new-feature, /review-pr, /fix-bug, /write-tests, /go, /catchup — consistently and not selectively. Contribute Skills Library Updates — at project close so the next team starts from a higher baseline. Feed lessons back to the shared skills — ea-standard, api-standard, and domain skills — via concrete LESSONS_LEARNED.md entries. Understand, apply and continuously Improve token usage - Improvise and fine tune token usage with AI- workflow and Incorporation of new skills Required Qualifications 4+ years of professional software engineering with production ownership. Fluency in the tech stack: .NET 8 / C# or TypeScript, EF Core / Drizzle, PostgreSQL, Cockroach DB, React.js, Python, REST + OpenAPI. Strong code‑review judgment and the ability to recognize subtly wrong AI output. Comfort operating as a director of agentic tools rather than sole author — and the discipline to verify, not trust. Solid testing discipline and familiarity with CI‑based quality gates. Bachelor’s degree in computer science, engineering, or a related field, or equivalent experience. Preferred Qualifications Hands‑on experience with Claude Code or a comparable agentic coding environment. Experience maintaining AI context and guardrail files (CLAUDE.md‑style) in a shared codebase. Domain exposure relevant to the pod (collision repair operations, fleet, insurance/DRP, or analytics). Track record of contributing reusable patterns back to a shared library. Core Competencies Director’s mindset — sets direction for AI tooling and evaluates output critically rather than accepting it at face value. Engineering rigor — holds a high bar on standards, test coverage, and code review regardless of whether code was human‑ or AI‑authored. Disciplined focus — scopes sessions tightly, commits frequently, and resets cleanly rather than pushing through degraded context. Learning instinct — captures what works and what does not and systematically feed it back to improve the team’s shared baseline. Collaborative accountability — works within the pod model and does not treat AI speed as an excuse to skip processes. Strong team player - comfortable working and partnering with cross functional team to continuously learn at high pace, build and continuously improve #J-18808-Ljbffr RedStream Technology