The Challenge

In modern technical hiring, software engineers face two critical barriers:

The ATS & Semantic Skills Graph Filter (Gate 1): Automated parser algorithms (Ashby, Greenhouse, Eightfold.ai) ingest resumes into vector co-occurrence graphs. Unstructured formats, multi-column tables, and missing technical primitives result in silent auto-rejection before human review.
The 6-Second Recruiter & Engineering Manager Scan (Gates 2 & 3): Human reviewers scan resumes in an F-pattern within 6 to 10 seconds. Generic bullet points (“responsible for backend services”), missing baseline metrics, and overused AI filler words ("spearheaded", "orchestrated") trigger immediate dismissal.
Typography & PDF Ligature Corruption: Traditional Word-exported PDFs and misconfigured LaTeX templates often corrupt Unicode ligatures (fi, fl, ff), causing ATS parsers to read words like efficiency as e ciency.

Architectural Solutions & System Pipeline

[Raw Candidate Text / Markdown / JSON]
                  │
                  ▼
   ┌──────────────────────────────┐
   │  75+ Rule AST Regex Linter   │ ──(Violations)──> [Actionable Rewrite Table]
   │  & Banned Words Dictionary   │
   └──────────────┬───────────────┘
                  │
                  ▼
   ┌──────────────────────────────┐
   │ Google XYZ & Harvard STAR-S  │ ──(Transform)──> [6-Dimension Quantified Bullets]
   │    Transformation Engine     │
   └──────────────┬───────────────┘
                  │
                  ▼
   ┌──────────────────────────────┐
   │ 4-Persona Hiring Simulator   │ ──(Stress-Test)──> [100-Point Hiring Scorecard]
   │ (ATS, Recruiter, EM, Bar-R)  │
   └──────────────┬───────────────┘
                  │
                  ▼
   ┌──────────────────────────────┐
   │ Typst / LaTeX Render Engine  │ ──(Compile)──> [Sub-50ms ATS-Hardened PDF]
   │ (Native CMap Unicode Tables) │
   └──────────────────────────────┘

1. 75+ Rule Regex Linter & AI-Slop Detection Engine

Constructed a rule-based AST analysis engine incorporating 75+ specialized linting rules that detect:

Weak passive duty phrases and missing action verbs.
Overused LLM hallucination markers and empty corporate clichés.
Banned visual elements (progress bars, rating stars, photo headers, multi-column tables).
Unquantified accomplishments lacking baseline comparisons.

2. Harvard / Google XYZ Transformation Engine

Automates the formulation of bullet points into the rigorous Google XYZ accomplishment model:

$$\mathbf{\text{Action Verb}} + \mathbf{\text{Technical Tool / Architecture}} + \mathbf{\text{Mechanism / Scope}} \longrightarrow \mathbf{\text{Quantified Impact / Metric}}$$

Every bullet point is anchored across at least one of the 6 fundamental engineering dimensions:

Latency & Compute: p95/p99 latency (ms), throughput (RPS), memory footprints.
Scale & Traffic: Active users (DAU/MAU), events processed/sec, data volume (TB/PB).
Reliability & Quality: SLA uptime (99.99%), test coverage (%), MTTR reduction.
Cost Efficiency: Cloud infrastructure spend reduced ($/yr or %), egress cuts.
Developer Velocity: Build/test cycle acceleration, automated CI/CD gating.
Scope & Team Impact: Squad size mentored, client accounts onboarded.

3. Multi-Persona Hiring Committee Simulator

Implements a 4-persona concurrent evaluation matrix:

Persona 1 (ATS Parser): Validates single-column flow, skill co-occurrence density, and ISO date parsing.
Persona 2 (Technical Recruiter): Evaluates 6-second F-pattern scanability and right-aligned date gutters.
Persona 3 (Engineering Manager): Checks low-level architectural primitives and eliminates tutorial clones.
Persona 4 (Amazon Bar Raiser): Audits against 16 Leadership Principles and tests the candidate’s “30-Second Interview Defense”.

4. Next-Gen Typst & LaTeX Compilation Architecture

Developed automated compilation pipelines utilizing Typst and hardened LaTeX:

Instantaneous Compilation: Typst compiles complex, paginated career documents in < 45ms (vs. 2.8s for LuaLaTeX).
ATS Font Hardening: Native CMap Unicode font embedding guarantees zero character corruption or broken ligatures when ingested by ATS OCR engines.
Strict 1-Page Layout Math: Dynamic line-budget and spacing algebra prevents 1.1-page overflow traps.

5. 2026 Agent Skills Specification Integration

Packaged the entire intelligence engine as an installable Agent Skill compatible with Claude Code, Cursor, Antigravity, and Gemini CLI workflows, enabling AI pair programmers to audit and rewrite developer portfolios with zero context drift.

Tech Resume Expert | Career Intelligence & Resume Engineering Platform

Technologies & Tools