Farhan Shafee

Workflow automation / Systems engineering

Job Intelligence Engine

An engineering system for job discovery, normalization, matching against career evidence, and application workflows.

All work

Overview

I built this personal engineering system to bring job discovery, career evidence, application preparation, and progress tracking into one workflow.

Problem

Job searching across multiple sources fragments information and repeats the same evaluation and writing tasks. Application material also needs to stay grounded in actual career experience.

System

Manual imports, configured job-board discovery, and Outlook discovery feed normalized job records. The Career Truth Vault stores structured career facts and source references for requirement matching and application drafts.

Technical highlights

  • Career facts, source references, and verification states connect matching to application material.
  • Approved résumé versions are immutable; application submission remains external and user-confirmed.

Architecture

The React/TypeScript application uses Next.js-style App Router conventions through vinext and Vite. Source configuration targets Cloudflare Workers; Cloudflare D1 and Drizzle support persisted jobs, career evidence, assessments, and tracking.

Normalization, deduplication, and provenance precede matching. Selected jobs feed application preparation and tracking; Outlook also produces workflow suggestions for review.

Job inputs become normalized records. Those records and approved Career Truth Vault evidence feed matching. A user selects jobs for application preparation and review, while the tracker records confirmed external submissions.
Conceptual discovery and preparation path. Outlook also supplies tracker suggestions for review; discovered jobs are not automatically submitted.

What I built

  • Greenhouse, Lever, and Ashby public-feed adapters; JSON-LD job extraction; and LinkedIn public-page/import parsing.
  • Microsoft Graph Mail.Read integration for discovering job, recruiter, and application messages.
  • Evidence-linked matching, recruiter-message templates, optional cover-letter templates, screening-answer assistance, and résumé/application packets.
  • Résumé review and DOCX export, plus a tracker for stages, notes, follow-ups, outcomes, and event history.

Engineering decisions

  • Application assistance uses deterministic templates grounded in career evidence; requirement scoring is deterministic as well. The current workflow does not use a language model.
  • Retain evidence references and gaps. Unresolved personal screening questions require user input rather than an invented answer.
  • Run discovery against selected inputs and configured sources; do not treat a source adapter as unrestricted background crawling.

Security / privacy considerations

The Outlook connector requests read-only Mail.Read access; job relevance is filtered by the application, not by a job-only mailbox permission. Résumé approval and export use review and verification gates. The tracker requires confirmation of an external submission before marking an application Applied.

Current state

An active personal engineering system for my own job-search workflow, with assistive preparation and user-controlled tracking.