Visiting Engineer: Live Selling Copilot

Live commerce is bottlenecked by host capacity and operational chaos.

Great sellers win on personality and community, but they still have to multitask like air-traffic controllers: managing chat, pacing listings, handling inventory, answering niche questions, and doing post-show follow-up.

This role is to build a LiveOps Copilot: a human-first AI producer that runs backstage for live commerce sellers, keeps the host front and center, and automates the operational work around the show.

• A backstage producer cockpit that clusters chat, drafts or sends guardrailed replies, and surfaces the highest-value audience questions in real time.
• A listing and inventory control towebaer that helps hosts manage sequencing, pinned items, bundles, and oversell or stale-listing risk during the show.
• A real-time product research layer that answers unknown questions from seller notes, listing data, historical Q&A, and trusted category sources.
• Multilingual translation and localized reply support so hosts can engage cross-border audiences without losing flow.
• A lightweight post-show repurposing workflow that turns streams into reusable clips, Q&A moments, and listing assets.
• Build the end-to-end real-time system across chat ingestion, retrieval, guardrails, operator UX, listing controls, and post-show outputs.
• Work with one of our large LPs as your build/design partner to identify the right pilot category, host set, data path, and success metrics for the MVP.
• Design the automation ladder from draft support to host-controlled agentic replies without undermining authenticity or seller trust.
• Make decisions about latency, guardrails, disclosure, and escalation so higher-autonomy features are safe and useful.
• Work with AI Fund's build team on product wedge, technical risk, and the path from pilot to a broader live-commerce OS.
• Strong full-stack execution, especially on real-time systems, operator workflows, and AI-assisted products.
• Demonstrated experience building GenAI products with retrieval, tool use, guardrails, and production feedback loops.
• Product judgment for human-in-the-loop automation where the right answer is usually to remove busywork rather than replace the human outright.
• Ability to move quickly with external design partners and translate ambiguous feedback into working product decisions.
• You actively use AI coding assistants and modern GenAI tools as part of your build workflow.