A research workflow you can inspect while it runs.
Project brief
Turned opaque AI research into a workflow with live state, manual checkpoints, reruns, citations, and exportable reports.
Hiring signal
Senior ownership across product framing, scope, and shipped interface quality.
Engineering read
Focus on constraints, architecture choices, external systems, and data flow.
A guided multi-agent research assistant. You give it a goal, a planner breaks the work into stages, a research router grounds each step in web or reasoning, and a writer/critic loop turns the result into a cited report on local Ollama inference.
Most AI research demos are just one opaque model call. You can't see what's happening mid-run, you can't pause it, and you end up paying a hosted API for work that could run locally.
I broke the pipeline into five visible stages — planner, research router, researchers, writer, critic — and built the frontend around a React Flow graph so each step stays visible while it runs. Server-sent events keep the UI in sync with real state instead of a fake progress bar.
Project method
The starting requirement was not another chat interface; it was a transparent research workflow where a user could see how an answer was being planned, grounded, rewritten, and critiqued. That pushed the scope toward inspectable stages, manual checkpoints, rerun states, and citations rather than a single prompt-response screen.
Next.js App Router handled the product shell and streaming API routes, React Flow made the agent graph legible, and Material UI provided dense controls quickly. Ollama was chosen deliberately so the core promise could be local inference with no paid model dependency, while Server-Sent Events kept the run graph synchronized without the complexity of a bidirectional socket protocol.
The MVP was defined as one complete goal-to-report loop: planner, research router, editable researchers, synthesizer/writer, critic, and citation-aware output. Features that did not prove that loop, such as persistent multi-user memory or configurable remote model pools, were left for later.
The project moved from a linear orchestrator to a checkpointed pipeline. Once planner output existed, the next critical step was making stale research visible and rerunnable before synthesis, so the UI evolved around graph state, pause/continue controls, and review modals instead of a passive progress indicator.
Takeaway
The result is a pipeline you can follow: planner → router → researchers → writer → critic, with rerun badges for stale outputs, pause/continue controls, and no paid API in the loop because inference runs on Ollama.
Evidence
Open channel
Architecture, AI workflows, design systems, and interfaces that need to hold up after launch.