PathfinderAI
A proposed B2B2C recruitment and skill-verification platform that replaces static resume screening with adaptive roleplay interviews and job simulations, giving employers competency scorecards and candidates actionable practice feedback.
The Challenge & Core Problem
High-Level Briefing
The proposal identifies a trust and verification gap in recruitment: resume keyword matching can miss practical capability, while candidates often receive little guidance after rejection and employers have limited time for individualized assessment.
"Design a scalable way for employers to assess role-relevant reasoning and workplace skills while giving candidates a safe environment for practice and constructive feedback."
Hard Timebox & Rules
- The proposed system uses adaptive, job-description-based interviews and situational simulations rather than resume keywords alone.
- A multi-agent workflow is proposed to generate contextual follow-up questions and evaluate responses against employer-provided rubrics.
- Speech-to-text is planned as a mocked demonstration integration for the Phase 2 prototype; the core workflow is text-based.
Open Risks & Pressure Factors
- Candidate and employer needs must be served without making hiring decisions autonomously.
- The proposal identifies fairness, data privacy, model costs, and safeguards against unbounded agent questioning as design constraints.
Leadership & Personal Contribution
Team Member
Real-Time Architectural Dilemmas Under Clock Pressure
The Technical Solution
PathfinderAI proposes adaptive interview and job-simulation sessions that generate employer-facing competency scorecards and candidate-facing growth roadmaps. Candidates can also use the platform to practice without sharing a profile with an employer.
Data Flow & Architecture
The proposed LangGraph workflow ingests job descriptions, competency rubrics, resumes, and candidate responses. An Interrogator Agent generates context-aware questions and pushback; an Evaluation Agent compares responses with employer-provided rubrics. The system proposes separate outputs for HR and the candidate. Whisper-based Sinhala and Tamil support is described as a localization feature, with speech-to-text mocked for the proposed Phase 2 demonstration.
Critical Data Pipelines
The Retrospective: What Worked & What Needs Work
What Worked
- ✓The proposal identifies distinct needs for recruiters and candidates and gives each a separate output.
- ✓It describes an adaptive multi-agent workflow rather than treating a general chatbot as the assessment.
- ✓Human oversight, bounded interview turns, privacy controls, and bias mitigation are included as responsible-AI safeguards.
What Still Needs Work
- ✕PathfinderAI is documented here as a submitted proposal; no completed product, deployment, or judging outcome has been reported.
Hard-Won Engineering Takeaways
Adaptive Skill Verification with Human Oversight
PathfinderAI uses contextual roleplay to explore candidate reasoning, then turns the session into separate employer and candidate outputs while leaving final hiring decisions with people.