SiftHQ

The Problem

Hiring is broken at scale. When you're evaluating dozens of candidates through take-home assignments, the manual overhead kills velocity — scheduling, tracking submissions, reviewing, scoring, and coordinating across hiring managers.

At every company I've worked at, the process was the same: spreadsheets, email threads, and too many people doing too much coordination for too little signal.

What It Does

SiftHQ automates the assignment pipeline end to end:

• Assignment scheduling — candidates get assignments with deadlines, automated reminders, and submission tracking
• Submission evaluation — AI-assisted scoring with configurable rubrics, so reviewers focus on signal, not boilerplate
• Decision dashboards — hiring managers see candidate progress, scores, and comparisons in one place

The goal isn't to replace human judgment — it's to remove everything around it so the team can focus on the actual evaluation.

How It's Built

SiftHQ runs on the same stack I use for everything: React + Vite on the frontend, Supabase for database and auth, and Cloudflare Pages for hosting. The AI evaluation layer uses Gemini via Cloudflare Workers.

The admin interface is password-protected (no user accounts needed — this is an internal tool), and the candidate-facing flow is a clean, minimal experience designed to reduce drop-off.

What I Learned

Building SiftHQ taught me that the hardest part of AI-assisted evaluation isn't the AI — it's designing rubrics that produce consistent, useful scores. The model is only as good as the criteria you give it.

It also reinforced something I've seen in product leadership: the best tools disappear into the workflow. SiftHQ works because hiring managers don't need to learn a new tool — they just see results.