VOL. III · APPENDIX R · RECRUITERSŠimonre: your role · 2026
an entry, in lieu of a cold email

Hi.
One moment.

Paste a job description. Get my résumé and cover letter, tailored to the role, in seconds. Every detail is drawn from real experience and genuine motivations — no generic templates, no hallucinations.

I’m not currently looking — I just like building demos. If the role is interesting, I’ll read it.

R.03 →Paste a role to get started
Šimon Podhajský, seen through a café window in Prague
Fig. R-1. Šimon Podhajský. Head of AI. Prague.
simon.podhajsky.net /recruit§ R.01 — the portalp. 1 / 6
in four short movements

The method, alphabetized.

Four steps, each with a footnote. None of them are surprising; the surprise is that they all actually run against a real database.

re: how the portal works§ R.02 — the methodp. 2 / 6
I.
Intakesee also: § R.03

You enter your email and paste the job details (or a link). The portal infers your company from the email domain.

II.
Alignment checksee also: values, role-fit

Two checks run in parallel — company-values fit and role-fit (skills, seniority, domain). If something’s off, you’re told upfront, not after.

III.
Curation, not inventionsee: hallucinations, none

The system researches the company, then queries a vector database of real experiences and selects the most relevant ones. Nothing is fabricated.

IV.
Vended packet/dossier

Résumé and cover letter, streamed live, ready to forward. Reviewable, exportable, dated.

fill in only what’s needed

Share the role.

A short form. The portal does the rest. Either paste the description, or drop a link — whichever is faster for you.

re: your opportunity§ R.03 — the dossier-formp. 3 / 6
↳ company inferred from domain
↳ we’ll infer once you fill in the e-mail.
streamed live · curated, not generated

Your dossier.

Results will stream in here once you submit the role. Résumé, cover letter, and research snapshot update live.

re: live output§ R.04 — the dossierp. 4 / 6
awaiting submission

The page is, for now, blank.

Paste a role above. The dossier will fill in below, in roughly twenty-five seconds, including the company research.

cross-references for the curious

Footnotes.

Six entries. Hover or focus any row to reveal its note. They double as a cheat-sheet for how I think about this kind of system.

re: questions, frequently asked§ R.05 — appendix & footnotesp. 5 / 6
AArchitecturesee: stack, observable

A Next.js frontend collects the role details and streams results via SSE. The FastAPI backend first researches your company using Tavily web search, then runs two alignment checks in parallel — company-values fit and role-fit (skills, seniority, domain) — using structured LLM extraction. If alignment passes, it queries a PostgreSQL + pgvector database of real experiences via embedding similarity, selects the most relevant ones, and streams a tailored resume and cover letter through OpenAI. Everything is stored and observable via Logfire tracing.

graph LR
  A[Next.js Frontend] -->|SSE| B[FastAPI Backend]
  B --> C[Tavily Web Search]
  B --> D[Alignment Checks]
  D --> D1[Company-Values Fit]
  D --> D2[Role Fit]
  B --> E[PostgreSQL + pgvector]
  E -->|Embedding Similarity| F[Experience Selection]
  F --> G[OpenAI Generation]
  G -->|Stream| A
  B --> H[Logfire Tracing]
CCredentials, hallucinated?see: no

No. The system curates from a database of real experiences, skills, and projects. It selects and emphasizes what's relevant — it doesn't invent. In the future I'll add additional guardrails to detect even embellishment.

EExpectations, talking insteadsee: yes please

No. This is a practical demonstration of full-stack and AI engineering skills. If it makes you curious enough to reach out, it's done its job.

DData, what happens to it/policy

Submissions — your email, company, and job description — are stored so I can review what roles come in. Your data isn't shared or sold.

FFlagged by alignment, now what/method § II

The alignment check protects your time too. If the role isn't a strong fit — wrong domain, seniority mismatch, values misalignment — the system is transparent about it upfront rather than wasting your time with a generic packet.

WWhy build thissee: practical demo

Sending generic résumés felt like a missed opportunity to show what I can actually build. Also: I think it’s funny.