context
Turning research evidence into reviewed opportunities
Researching small engineering firms generated useful evidence but no consistent path from discovery to a reviewed opportunity and personalized contact draft.
problem
Replacing duplicate state and opaque AI outreach
Independent scripts created duplicate state, inconsistent statuses and limited visibility into the funnel. AI-generated outreach also required a hard human-review boundary.
approach
One traceable workflow with explicit approval gates
I designed a Python CLI and Flask dashboard around one canonical dataset. Agents support company research, weighted scoring, reports and draft generation. Event logs, status synchronization, validations and explicit approval gates maintain traceability.
results
A shared source of truth covered by automated tests
The system now exposes funnel KPIs across terminal and web interfaces and is covered by 54 automated tests. No external contact is sent without manual review.
next
Improving evidence quality and conversion measurement
Add stronger evaluation of research quality, improve deduplication across corporate entities and formalize conversion metrics over a longer operating period.