AI Development & Support Engineer

How to Apply Provide the following required materials: Resume Cover Letter – In your application, you must submit a cover letter. Please provide brief, concrete examples for the following questions. We value specific technical details, tool names, and honest assessments of where technology succeeds and fails. Describe a project where you used an AI tool (e.g., ChatGPT, Claude, Cursor, GitHub Copilot, etc) to write or significantly speed up your coding. What were you building, what was your specific approach to instructing the AI, and where did the AI fall short? All LLMs have specific coding quirks and repetitive errors. Name the specific model you use most (e.g., Opus, GPT-5.x) and describe a highly specific, recurring syntax or logic error it makes that drives you crazy. How do you work around it? If a hot new open-source AI model gets released tomorrow and you want to test it out locally on your own laptop or machine (not just using a web browser), how would you set that up? What specific software, tools, or frameworks would you use to get it running? Location Your home office or our Waltham, MA HQ or hybrid Qualifications Hard Requirements Solid grasp of agent architectures, prompt engineering, and MCP or tool-use patterns Hands-on experience with local/open-source LLMs: you've run them, tuned them, deployed them Comfortable working in Linux environments and debugging at the command line Strong communicator: you can explain what's broken and why to both engineers and non-engineers This is not a role for someone who has tried out some youtube tutorial videos on their own. You need to have some real experience implementing this. Helpful Background 2+ years Python development — real shipped and used code Experience with RAG pipelines or vector databases Familiarity with REST APIs and integration patterns Docker comfort Background in storage systems or HPC environments (this helps more than you think) Working Style You can context-switch between deep development work and supporting a customer mid-debug You're comfortable being the person who figures things out when there's no playbook yet About Starfish Storage Starfish Storage is an established company that maintains a startup culture mindset. We cater to a diverse client base, serving industries such as life sciences, artificial intelligence, and research computing. Our technology is critical to modern organizations, helping customers organize, manage, and move the tens of billions of files central to their operations. Most AI work right now is happening in the cloud, with managed APIs, and relatively clean data. Some of our customers can't do that: they're operating in classified or regulated environments with filesystem metadata spanning billions of files. The AI problems here are genuinely hard, the deployments are genuinely constrained, and the work actually matters to science and research. If you want AI work that's genuinely constrained, genuinely consequential, and can't phone home to OpenAI — this is it. Job Overview Starfish manages petabytes of unstructured data for national labs, research institutions, fintech, and pharma, the kind of data that doesn't fit neatly into databases or cloud buckets. We're building an AI-native layer on top of that: MCP servers, agentic workflows, local LLM deployments, and tools that let researchers and engineers actually talk to their storage infrastructure. We're tackling hard questions too: should an agent be permitted to move 2PB of data autonomously? How do you govern that? In this role, you'll build and ship MCP servers, deploy and tune local LLMs, write custom agents for real customers with real petabyte-scale problems, and get them running in environments where "just use the API" isn't an option. You'll also support customers when things don't work: which means you need to actually understand what you've built. Responsibilities and Duties include: MCP & Agent Development Build and expand MCP servers that expose Starfish's data management capabilities to LLM-powered workflows Develop intelligent agents that can reason over storage metadata at scale — including local LLM deployment and configuration using vLLM, Ollama, or similar Create workflow automation and chat-based interfaces for querying and acting on Starfish data Build proof-of-concept implementations and turn them into production-grade features Customer-Facing Technical Work Support customers through installation, configuration, and troubleshooting of AI feat