Hidden Gems Talent: Architecting an AI Recruitment Pipeline

Role: AI Team Lead (Intern)
Location: Singapore / Melbourne
Timeline: Jun 2025 – Aug 2025

During the Monash Teamwork Internship, I led a cross-functional team in collaboration with Hidden Gems Talent—a recruitment agency in Singapore/Malaysia—to solve a critical operational bottleneck: candidate screening fatigue. We architected a complete, end-to-end AI-powered evaluation workflow that fundamentally transformed their hiring pipeline.

The Architecture & Stack

To ensure the solution was maintainable by the client post-handover, we built a fully automated workflow using low-code orchestration integrated with advanced LLMs:

• Workflow Orchestration: n8n (Drag-and-drop pipeline logic with custom JavaScript nodes for data formatting).
• AI Evaluation Engine: OpenAI API (Configured for multi-metric scoring and rigorous prompt instructions).
• Data Layer & Comms: Google Sheets API, Gmail SMTP, and Calendly integrations.

Core Technical & Leadership Achievements

1. Intelligent Candidate Evaluation Logic

The system doesn’t just parse resumes; it evaluates qualitative candidate responses. I engineered the LLM prompts to score applicants based on specific metrics: response clarity, proper use of the STAR method, grammar, and level of professional ownership.

2. Dual-Filter Shortlisting & AI Detection

To ensure candidate quality and authenticity, the pipeline implements a strict dual-filter. It requires a high qualitative score (≥ 7) while simultaneously applying ML classification to generate an “AI-likelihood” score. Only candidates who score highly on merit and low on AI-generation (< 5) pass the filter.

3. End-to-End Automation & Dynamic Actions

The n8n workflow reads incoming applications directly from Google Sheets. Based on the AI evaluation, it dynamically routes actions without human intervention:

• Shortlisted Candidates: Receive a next-steps email with an automated Calendly booking link.
• Rejected Candidates: Receive constructive, highly personalized feedback generated by the LLM based on their specific evaluation shortfalls.
• Internal Stakeholders: The CEO automatically receives a daily email digest summarizing the top-tier candidates.
• Database Updates: The system writes the final score, AI rationale, and application status directly back to the master tracking sheet.

The Impact

As the AI Team Lead, I was responsible for architecting this system from concept to client deployment. By implementing this modular agent pipeline and establishing reproducible experimental workflows, our team successfully reduced manual candidate screening by 95%. This project stands as a prime example of leveraging API orchestration to drive massive, measurable business efficiency.