AI Resume Pipeline

Overview

The AI Resume Pipeline is the flagship enterprise AI system I built at Jacobs Engineering. Deployed in Palantir Foundry, it automates the complete lifecycle of candidate resume processing — from document ingestion through structured AI extraction to standardized document generation.

The system handles PDF and DOCX resumes up to 23 pages and 60,000+ characters, using Claude Sonnet with a chunked extraction strategy to reliably parse complex documents. It extracts 65+ projects per resume, applies PII governance, enriches data against internal HR datasets, and produces formatted outputs ready for downstream hiring workflows.

Problem

HR teams were spending 4+ hours per resume on manual processing. Each document required careful review, data extraction, formatting corrections, PII handling, and enrichment before entering hiring workflows — a bottleneck that did not scale with hiring volume.

• Large document sizes (60K+ characters) exceeding manual review capacity
• Inconsistent formatting across resume sources and file types
• Manual data extraction prone to errors, delays, and inconsistent outputs
• PII governance requirements across every processing stage
• HR data enrichment needed before final document generation
• No automated path from raw upload to standardized enterprise resume

Solution

I designed and built a fully automated pipeline orchestrating ingestion, parsing, AI extraction, data protection, enrichment, and document generation within Palantir Foundry. Key engineering decisions centered on chunked extraction for large documents, few-shot prompt engineering for consistent structured outputs, and PII scrubbing integrated before any enrichment step.

• Automated ingestion from HR document sources with format detection
• Multi-format document parsing for PDF and DOCX inputs
• Claude Sonnet LLM extraction with chunked processing for 60K+ character resumes
• Few-shot prompt engineering for consistent structured data extraction
• PII scrubbing workflows enforced before HR data enrichment
• HR data enrichment against internal enterprise datasets
• Automated document generation with standardized enterprise templates
• Scheduled orchestration for continuous batch processing

Technology Stack

• Palantir Foundry
• Python
• Claude Sonnet
• LLMs
• Prompt Engineering
• Data Pipelines
• Automation

Engineering Highlights

• Chunked document processing strategy for 60K+ character resume files

• Few-shot prompt engineering achieving consistent structured extraction

• Multi-format support across PDF and DOCX with unified parsing pipeline

• PII governance workflows integrated at every processing stage

• HR data enrichment pipeline connecting AI outputs to enterprise datasets

• Automated scheduling and orchestration for production batch workloads