Startup research ops lead — pilot proposal

Role and vacancy summary

A growth-stage fund needs a research operations lead to deliver a repeatable weekly shortlist: ingest new companies, rank fit against a thesis, attach evidence links, and hand reviewers a package they can approve in under an hour. The role is not “build another chatbot.” It is “ship a workflow that survives audit, budget caps, and a skeptical investment committee.”

How AI OS fits

AI OS can act as the control plane for that workflow: curated intake, RAG-backed context packs, deterministic gates before anything is published, and human approval on scope changes. Public portal pages stay read-only; the operator dashboard holds orchestration, effectiveness notes, and review checkpoints.

What we would not promise in a first pilot: automatic deal decisions, guaranteed coverage of every niche data vendor, or unconstrained agent loops without success tests.

Curated capability snapshot (public tier)

• 2000+ regression tests across the core product surface.
• 17+ product modules (orchestration, RAG, OSINT wrappers, public portal, effectiveness, and related surfaces).
• 40+ public portal sections with honest empty states when relays are unavailable.
• Gated factory pattern: narrow tickets, isolated workers, budget caps, and deterministic checks before user-facing output.

Value for this client

• Repeatable weekly brief instead of ad-hoc copy-paste across spreadsheets and chat threads.
• Evidence-linked summaries reviewers can challenge line by line.
• Bounded automation: agents run only where a success test exists; everything else stops for human input.
• Clear separation between public marketing pages and operator-only execution surfaces.

Honest gaps

• AI OS is not a CRM, ATS, or portfolio system of record. Exports and handoffs to your existing tools remain your integration work.
• Startup coverage quality depends on configured sources and reviewer discipline; we do not guarantee “every company on the internet.”
• Novel one-off research questions may still need senior analyst time; factories accelerate repeatable slices, not magic answers.
• Full orchestration and write paths require authenticated operator access; the public portal alone cannot run the pipeline end to end.

Recommended next step

Run a two-week bounded pilot on one thesis vertical: define inputs, acceptance checklist, reviewer roster, and weekly output format. Success = reviewers can reproduce the brief without asking “where did this claim come from?” Failure modes are documented up front, not hidden behind a demo button.