Yashdeep Singh

Case study, AI agent for a Berlin water-tech startup

Research email agent

Finds the right people at public water authorities, checks every contact against its source and prepares the outreach list.

Role
Built it end to end
Stack
Claude with web search, Python
Output
Contact database and PDF

The problem

Authorities list a general inbox. The people who decide are hidden in team pages.

How I built it

Target listStart from a list of authorities with their general contact details
ResearchClaude with web search reads each office's team pages and finds the groundwater department
StructurePython writes contacts and organisations into a linked database with IDs
Prepare outreachSalutations and a shareable PDF list, each contact linked to its source

Testing and what I fixed

Problem

Big batches mixed up names.

Fix

Three to four offices at a time, each contact linked to its source.

Problem

GDPR.

Fix

Opt-out status built into the database.

What I would do next

Automatic re-checks every few months.