AI Engineer

Putco Inc is hiring an AI Engineer for an onsite role in Des Moines, IA, focused on owning and hardening a closed, on-prem AI stack. You will connect the system to the Macola ERP and Putco’s catalog and OEM workflows, drive an upgrade to a ~240B-class multi-GPU NVIDIA workstation, and produce measurable demos and reporting for leadership.

Key Responsibilities

• Own and harden the closed on-prem LLM environment, starting with NVIDIA Spark, with a defined path to multi-GPU hardware for ~240B-class models (hardware such as 4× RTX 6000 Ada is preferred, with Western models such as Llama / Mistral families).
• Build Macola intelligence using read-only SQL/schema for RAG, including accurate report library work to replace legacy reports that are flawed.
• Support Excel to Macola BOM and article loaders with a human-approve write-back approach so AI does not silently update production data.
• Develop mothership connectors for the full putco.com catalog and OEM sources (Ford, GM, Nissan, Mopar), including attribution across wholesale, retail, web, and OEM, plus catalog generation in sequenced waves.
• Operate within a LAN-only, firewall-locked deployment using Open WebUI or an equivalent approach, with no leakage of Putco data to outside AI APIs unless an exception is locked and capped by James.
• Deliver Friday demos to James / Paul using real Macola questions and measurable milestones, avoiding slide-only updates.
• Hand off work cleanly from the current contractor SOW by documenting processes and system details so Putco is not dependent on a single person.

Required Qualifications

• Hands-on experience running open-weight LLMs on NVIDIA hardware (examples include Ollama, vLLM, TensorRT-LLM, or similar), with experience beyond cloud SaaS prompt engineering.
• Strong SQL skills and a willingness to work within a mature ERP schema (Macola or similar manufacturing ERP is a plus).
• Python (or equivalent) for loaders, RAG pipelines, report automation, and light internal UIs.
• A security mindset aligned to closed LAN, least privilege, and avoiding unsafe patterns such as pasting internal data into ChatGPT.
• Ability to ship and demo weekly; comfort with onsite collaboration in the Des Moines area with ops, engineering, and sales users.
• Clear English suitable for CEO-level updates.

Technology Stack

• NVIDIA Spark, NVIDIA hardware, Ollama, vLLM, TensorRT-LLM
• Llama, Mistral
• SQL, Python, RAG
• Open WebUI, RTX 6000 Ada

Benefits

• 401(k)
• 401(k) matching
• Dental insurance
• Employee discount
• Flexible spending account
• Health insurance
• Life insurance
• Paid time off
• Vision insurance

Success Criteria (First 90 Days)

• Spark (or successor) becomes the trusted Macola brain: ai_readonly plus schema-based RAG answering real ops/engineering questions with cited SQL.
• Core backlog reports ship as Excel plus email digests that James actually uses.
• BOM / article loader path is live with human approval gating.
• Written hardware and model plan for the ~240B upgrade, including pricing and sequencing (B&H / CDW class kit).
• Mothership Wave 0–1 underway with the full data plane connected and measurable progress beyond a single-SKU pilot.

Nice-to-Haves

• Manufacturing / automotive aftermarket / OEM documentation familiarity.
• Experience correcting bad legacy SQL report logic.
• Headless commerce / catalog feed work (Woo or similar) or PLM adjacency (Aletiq is a separate system and should not be blended without approval).
• Multi-GPU workstation build experience, including ConnectX and high-VRAM inference tuning.

How the Team Works

• They start with a closed system first, with cloud APIs only under a hard monthly spend cap set by James.
• Paul Elwell owns digital / site execution, while you own the Macola + on-prem intelligence layer and how it feeds the mothership.
• Aletiq PLM stays separate unless James expands scope in writing.

Role Details

Location: Des Moines, IA (onsite)

Compensation: USD 125,000 to 155,000 per year