Senior Applied AI Engineer - Remote Production ML

Who Is Vantaca

Powered by AI, Vantaca is a leading AI‑native community management performance platform that enables owners, operators, community management teams, boards, and associations to work smarter, faster, and with unprecedented insight. Vantaca is intelligent business operating software that leverages artificial intelligence to automate routine work, surface actionable insights, and help our customers increase revenue, efficiency, flexibility, and control.

We are building a world‑class Applied AI practice inside Vantaca's Applied AI team. We need someone who can ship production‑grade ML and LLM systems for our Implementation and Client Enablement teams. This is not a prompt engineering role or an AI exploration sandbox; you will build systems that are evaluated, deployed, and observed – owning the gap between “interesting model” and “thing that reliably runs in production.”

You will partner with Implementation PMs, Solution Consultants, and Client Enablement Specialists to identify the highest‑leverage problems and ship tooling that removes friction across the client lifecycle. The work is high‑trust and high‑autonomy – you own your problem space end to end.

• Our Implementation and CE teams have a validated backlog of high‑value AI builds – risk surfacing, workflow intelligence, client coaching, configuration assistance – and no dedicated engineering resources to execute on them. You change that.

• Design and ship ML and LLM systems spanning supervised models that predict and rank, retrieval and generation systems that draft and summarize, and agentic workflows that act on internal data.
• Build evaluation infrastructure alongside every system – define success criteria before writing code, measure whether the system worked, and catch regressions before users do.
• Architect retrieval‑and‑generation, retrieval, and context engineering patterns that let LLMs operate reliably on internal knowledge and production data.
• Reason rigorously about modeling choices – label definition, leakage, time‑aware splits, calibration, precision‑at‑k vs AUC, when a heuristic baseline beats a model.
• Work directly in Databricks and Unity Catalog – understand the operational data, write the SQL, and build systems that act on it.
• Own deployment and monitoring for everything you ship – feature drift, outcome tracking, LLM eval regression, retraining cadence, rollback paths.
• Treat data governance and access scoping as design constraints, not afterthoughts.
• Maintain versioned, traceable LLM workflows – prompts and context patterns that are reusable, not one‑off.

• Production experience shipping both classical ML and LLM systems – strong opinions on when to use which.
• An eval‑first mindset – you don’t trust a system you haven’t measured, and you build the measurement before the model.
• Fluency in a data warehouse environment – SQL, time‑aware feature engineering, leakage discipline.
• Production scars – you’ve watched a model degrade in the wild, seen a label loop bias itself, caught an LLM provider regression with the prompt unchanged.
• Cost intuition – you can napkin‑math the unit economics of an LLM workflow before committing to it.
• Ability to scope work in partnership with non‑technical stakeholders, translating their pain into a buildable system.
• Comfort with distinguishing the business metric from the model metric, and arguing for the right one.

• You use AI tools (Claude, Cursor, Claude Code, or equivalent) as a core part of your daily workflow, not occasionally.
• Python proficiency as your primary build language for automation and scripting.
• Full‑stack range: comfortable building APIs, automations, integrations, and lightweight UIs without needing a separate front‑end resource.
• SQL and data fluency: you will work regularly in our data warehouse and need to understand and act on operational data directly.
• API integration experience: REST, webhooks, OAuth.
• RAG and retrieval system experience: chunking, embedding strategies, retrieval quality, hallucination mitigation.
• Prompt and context engineering: you understand why context boundaries matter and have a…