A Research Portal That Replaced Spreadsheets in 7 Days
The Challenge
DahliaTEQ is a three-person boutique recruiting firm specializing in executive search. They source candidates from think tanks, consulting firms, and academic institutions — tracking hundreds of professionals across multiple sources with complex scoring criteria.
Their process was spreadsheets, manual web research, and scattered notes. It worked for a while, but it didn't scale. They needed a centralized database with scoring, filtering, and the ability to track candidates across multiple job openings.
Off-the-shelf applicant tracking systems weren't built for this. Those tools handle inbound applicants. DahliaTEQ does proactive research — finding people who aren't applying anywhere. Traditional custom development quotes came back at $40-80K with a 3-6 month timeline. For a three-person firm, that's not realistic.
Our Approach
| Day | What Got Built |
|---|---|
| 1-2 | Core application — Next.js project with auth, PostgreSQL schema, candidate management UI, project and job structure |
| 3-4 | Scoring and filtering — multi-criteria scoring system, tier classification (A/B/C/D), advanced filters by source, role, tier, and location |
| 5-6 | Data pipeline — scrapers for 9 different sources (CFR, Asia Society, Kroll, and others), image scraping with S3 storage, import utilities |
| 7 | Polish, deploy to production on HostKit, client walkthrough and handoff |
The scrapers were the interesting part. Each source has a different structure, different data quality, different image handling. We built Playwright-based scrapers that could pull candidate profiles, normalize the data, and score them against active job requirements — all feeding into one clean interface.
The Outcome
What the client got on day 7:
- 609 scored candidates from 9 integrated sources
- 1,827 professionals in a centralized talent pool
- 5 active job requisitions configured with scoring criteria
- Profile images scraped and hosted on S3
- Advanced filtering — by source, tier, role, location
- Notes system for collaborative candidate evaluation
- AI-powered scoring with written reasoning for each candidate-job match
- Role-based access so the whole team could work simultaneously
| Metric | Value |
|---|---|
| Development time | ~7 days |
| Scored candidates | 609 |
| Total talent pool | 1,827 |
| Sources integrated | 9 |
| Codebase | ~7,500 lines |
Tech stack: Next.js 15, React, Tailwind CSS, PostgreSQL, Prisma, HostKit, MinIO (S3), Playwright, Claude Code.
Why This Worked
- The right tool for the job. No ATS on the market supports proactive research workflows. Custom was the only option — the question was how fast and how cheap.
- Scrapers compound value. Once the data pipeline existed, the client could re-run imports as new candidates appeared across their 9 sources. The tool gets more valuable over time.
- Small team, simple auth. Three users meant we could focus on the workflow instead of enterprise access controls. Right-sized architecture for a right-sized team.
- Ship the whole thing. Database, UI, scrapers, scoring, images, auth, deployment — all in one sprint. No phased rollout, no "MVP first." They got the complete tool on day 7.
