JobHunt Pipeline & Live AI Interpreter
Autonomous ATS discovery pipeline, zero-hallucination Chromium resume compiler, and low-latency FA⇄EN virtual-microphone audio bridge.
01 — The Operational Problem & Stakes
Senior remote roles on Greenhouse, Ashby, and Lever receive hundreds of applications within hours, requiring rapid role discovery and crisp track-specific resume positioning (AI Automation Track A vs. Full-Stack Web Track B).
Crucially, automated tailoring must never invent employers, alter employment dates, or contradict verified personal constraints — and generated PDFs must never spill awkward orphan lines onto a second page.
02 — System Architecture & Execution
The JobHunt Engineering Kit combines Python crawlers across 10 remote job boards (routed cleanly via a Cloudflare Worker relay) with an AST/regex truth-bank validator anchored to `80-owner-resume.md`: any LLM-tailored bullet that introduces an unverified employer, date, or metric is rejected automatically before PDF compilation.
Alongside the async written workflow sits an experimental real-time audio R&D rig wiring virtual PulseAudio/PipeWire sinks to Gemini multimodal streaming audio for low-latency Persian-to-English speech translation.
03 — Interactive Subsystem Topology (Click Any Node)
ATS Crawler + Cloudflare Relay
Fetches structured JSON/HTML postings across remote-first engineering boards and scores async/stack alignment.
04 — Key Engineering Decisions & Trade-Offs
Hard Fact-Locking Over Creative LLM Rewriting
A resume pipeline is only useful if every line survives deep technical due diligence; locking entities to the Vault truth bank made tailoring 100% trustworthy.
Automated Single-Page Spill Prevention
`scripts/render-resumes.mjs` measures rendered Chromium `scrollHeight` and fails immediately if typography or margins push even one line onto page 2.
Dedicated Track Separation (Never Muddy Positioning)
Hiring managers for AI Automation roles search different signals than Next.js/WordPress leads; maintaining distinct Track A and Track B artifacts maximizes relevance without inventing facts.
Turning Async Written Precision Into the Primary Wedge
Clear written specs, documented PRs, and fast paid work-sample trials outperform vague introductory calls for distributed engineering teams.
05 — Verified Build, Test & Runtime Proof
$ node scripts/render-resumes.mjs --portfolio-url https://hamidmehrani.pages.dev
✓ Track A (AI Automation Engineer): scrollHeight=644px <= 1040px -> 1 page A4 (38.7 KB)
✓ Track B (Web Developer Next.js/WP): scrollHeight=618px <= 1040px -> 1 page A4 (34.7 KB)
✓ Truth-bank consistency check against 80-owner-resume.md: PASS (0 unverified claims)