The feature that turns interview prep from 2 hours to 20 minutes

Every engineer I know who's been interviewing recently spends the same two hours before each call doing the same research.

Company blog. Engineering posts on Medium. Glassdoor. LinkedIn profiles of the team. Blind. The job description again, this time actually reading it.

You're looking for: what do they actually use (not what the JD says they use)? What's the culture signal in the way they write about themselves? What should you actually ask them?

Two hours. Per company. Times however many first-round calls you have this month.

EverCV's new company research feature does it in 20 seconds.

What it generates

Given a company name, a role title, and an optional job description, the /api/company-research endpoint calls Claude Haiku and produces a JSON dossier with five sections.

Tech stack with rationale. Not just "Python" — "Python, confirmed in JD requirements section and consistent with their open-source tooling on GitHub." Each item has a sentence explaining why it's likely, which tells you how much to weight it.

Culture signals. Observable things about how the team works, drawn from the JD language and the company's public reputation. A JD that says "move fast and own your work end-to-end" means something different than one that says "collaborate cross-functionally with stakeholders."

Questions to ask. This is the part I spent the most time on the prompt for. Bad interview questions are everywhere. "What does success look like in this role?" tells you almost nothing — every hiring manager has a stock answer for it.

Good questions reveal operational truth. The prompt explicitly bans the placeholders and asks for specifics grounded in what the JD actually said:

"Your JD mentions 'high velocity' — what does a typical deploy pipeline look like and how often do you ship to production?"

"What's the biggest technical incident you've had in the last 6 months, and how did you handle it?"

"Your stack includes both Kubernetes and Lambda — what drove that split, and is that intentional or historical?"

Red flags. If the JD lists 12 required technologies or uses phrases like "must be comfortable with ambiguity" without context, that warrants a closer look. The model flags them directly so you know what to probe.

Interview process estimate. Series B companies interview differently than FAANGs. A rough format estimate — screening call, technical, system design, offer — sets expectations before you go in.

The prompt design problem

The challenge with this feature is that Claude's training data isn't real-time. It knows a lot about Stripe, Google, and Cloudflare, but less about the 200-person B2B SaaS you're interviewing at next Tuesday.

The solution: prioritize the JD text over general knowledge, and be honest about confidence. The knowledge_caveat field in every response includes a one-sentence note on what's well-established versus inferred. "Tech stack based on JD; interview format is an estimate for a company of this size."

I'd rather an honest "I'm guessing about the interview format" than a confident hallucination.

Why this is a Pro feature

The other Pro features — cover letter, salary negotiation, warm intro — justify their value in outputs you send to other people. Someone else sees the cover letter. The negotiation email goes to a recruiter.

Company research is different. The value is entirely internal: you walk into that call better prepared. You ask a question nobody else asked. You don't get caught flat-footed on "do you know what we use for observability?"

That kind of preparation is hard to quantify directly, but it's the difference between an interview that feels like a conversation and one that feels like a quiz you're failing.

Using it in EverCV

The feature is in the dashboard under a new "Company research" section. Enter company name and role, optionally paste the JD, hit the button. The output renders inline: tech stack list, culture signals, questions numbered and ready to paste into your notes, red flags in red.

No copy-paste work between tabs. No reading Glassdoor threads looking for signal. The 20 minutes you save, you spend on the actual call.

EverCV is building toward a single place where your CV updates itself and every job application gets the same preparation level that previously only happened for dream jobs. Company research is the newest piece. Try it at evercv.io.