AI prototype · May 2026 - Present
NetworkMind
A two-person prototype for contact research and relationship intelligence.
A user can ask a question, search imported contact data, trigger public research where useful and receive ranked contacts with context.
Team
Two-person side project
My focus
Routing, retrieval, ranking and validation
Stack
Next.js, PostgreSQL, Prisma and Redis
Why we built it
Professional network research often starts with a contact export, incomplete notes and repeated browser searches. We wanted to see whether one focused workflow could make the process easier to follow.
How it works
- Import professional contact data.
- Interpret the user's question and select the required tools.
- Search contact records and perform public web research when it adds value.
- Rank relevant people and return an answer with supporting context.
My contribution
- Prompt routing and tool selection.
- Retrieval and public research logic.
- Contact ranking and fallback handling.
- Validation, testing and full-stack implementation.
Current state
NetworkMind is a working two-person prototype. It is not a commercial product or a production platform.
What I learned
The project changed how I think about AI applications. The model response is only one layer. The surrounding data flow determines whether the result is useful and trustworthy.