Minerva Foundry
CASE STUDY: Resumagic

122k LOC 255 Commits 8.3 Avg CC

Running the Job Hunt Like a Product

How I built a privacy‑first, AI‑accelerated CLI to strip away guesswork, create credible feedback, and make the process mine again.

8 min

Overview

Job hunts are brutal — slow feedback, unclear decisions, and too many unknowns. I’ve built products to solve tougher problems than this, so I decided to build one for myself: a workflow to identify and remove every blocker in the way.

“Treat the job hunt like a product: make feedback loops fast, decisions explainable, and privacy non-negotiable.”

Most job hunts are slow, opaque, and risky for privacy. Resumagic flips that script. Built in active development, it’s a deterministic, ethics-forward toolchain for executive-level candidates — pairing with AI as a collaborator and building on AI as the foundation. The result: faster iteration, explainable decisions, and full control over sensitive materials.

Who It’s For

  • Primary audience: Senior candidates and executives in tech, product, and adjacent fields.
  • Early adopters: Leaders who value privacy, speed, and clarity — and want to work with AI without ceding control of narrative or direction.

Core Product Principles

  • Human-led, AI-accelerated: Automate the repetitive; protect space for human taste, tradeoffs, and strategy.
  • Clarity with explainability: Every decision is paired with a clear “why” linked to specific edits or recommendations.
  • Privacy by default: Sensitive materials never leave my machine — unlike most resume optimizers, nothing is uploaded to third-party servers.
  • Fast loops: Small, measurable iterations; immediate feedback; no bottlenecks.

Core Features

Resumagic is more than a workflow — it’s a bundle of capabilities that directly address the hidden pain points of modern job hunts.

ATS-Optimized Resumes

  • What it does: Generates resumes in DOCX with layouts and formatting designed to parse cleanly across major Applicant Tracking Systems.
  • Why it matters: No more silent rejections from hidden format issues like headers, tables, or fonts.

Automated Keyword Extraction

  • What it does: Uses IR/NLP methods (TF-IDF scoring, semantic grouping, resume-injection checks) to surface a role-specific keyword checklist.
  • Why it matters: The system does the scanning; I focus on the narrative — what to emphasize, cut, or reframe.

Simulated Hiring Review Board

  • What it does: Six calibrated personas (HR, Technical, Design, Finance, CEO, Team) review a resume or cover letter before it’s sent. Fast mode for iteration, quality mode for deeper reads.
  • Why it matters: Anticipates reservations, exposes blind spots, and mirrors how real review boards read applications — but on demand and in private.

Read the deep dive: Inside the Review Board

Fast, Private Iteration Loops

  • What it does: A local-first pipeline with immediate turnaround (seconds, not days).
  • Why it matters: Cuts iteration cycles down from days to hours; candidates move faster and stay top-of-pile.

Impact Snapshot

In my own search, Resumagic has:

  • Cut job discovery → submission cycles from 1–3 days to under 2 hours.
  • Raised ATS readability scores from ~55% baseline to consistently 90%+, across third-party scanners.
  • Increased iteration velocity from 1–2 edits to 5+ high-quality loops per application.
  • Surfaced 2–3 role-specific blind spots per application through the simulated review board — feedback you’d otherwise only get after rejection.

Why It Matters

Resumagic is more than a personal productivity tool — it’s a proof-of-concept for how AI can support high-stakes, high-context processes without eroding trust, privacy, or human authority. It reflects the product mindset I bring to any challenge: focus on outcomes, build ethical guardrails in from the start, and design for both speed and clarity.

Path to Productization (high level)

Market signal. ATS is ubiquitous — ~98% of Fortune 500 companies use an ATS — and ATS-readiness tools are now table stakes for serious candidates. Meanwhile, the ATS software market is measured in billions annually, and the resume builder category is a sizeable, growing consumer market with a fast-growing AI subsegment. (Jobscan, Fortune Business Insights, WiseGuy Reports, Future Data Stats)

Competitive set (representative).

  • Jobscan — resume-to-JD “match rate” optimization with ATS-style scans (they recommend ~75%+ match). (Jobscan)
  • Resume Worded — ATS resume checker, AI rewrites, LinkedIn feedback. (Resume Worded)
  • Teal — resume checker plus job tracker and workflow tooling. (Teal)
  • VMock — university-focused AI resume feedback used across campuses. (careers.usc.edu)

Differentiation thesis.

  • Local-first privacy by default; explainable, deterministic loops (vs. opaque cloud uploads and black-box scores).
  • Multi-perspective, simulated review board that mirrors real decision-makers (HR, functional leader, finance, team) — not just keyword matching.
  • Built with AI and on AI: human-in-the-loop narrative control + AI-native pipeline for speed and quality.
  • ATS-safe document generation (DOCX pipeline) that prevents silent format failures while preserving a professional narrative.

Go-to-market (sketch).

  • B2C Pro (subscription): individual candidates who value privacy, speed, and explainability.
  • B2B2C Cohorts: universities, fellowships, outplacement firms, incubators/accelerators (review-board libraries per industry/seniority).
  • Executive tier / self-hosted: privacy-sensitive segments (healthcare, defense, academia) with local-only processing and policy controls.