FTJA:
Fine Tune Job Agent
FTJA is a Fine Tune Job Agent that semantically analyzes resumes, portfolios, natural language filters, and job postings to categorize suitable positions and provide justifications. By building an advanced RAG pipeline, it filters out irrelevant postings with 95%+ accuracy and has reduced evaluation costs to just $0.01 per query.
95% of the job search funnel is entirely irrelevant.
During product discovery, we identified a massive inefficiency: job seekers spend the majority of their time filtering through noise. Existing platforms rely on superficial keyword matching, failing to understand actual competence.
Furthermore, current AI matching tools are often "black boxes" that force users off their preferred platforms and offer no transparency over the matching logic. We needed to build a scalable prototype combining LLMs and Vector Search to evaluate true semantic fit—putting an intelligent, explainable workflow directly inside the job boards users already use.
Fine Tune Job Agent
Natural Language Filtering
Moving beyond simple filters and keyword-based searches. Users define non-negotiable deal-breakers (e.g., visa status) and flexible preferences (e.g., "0-to-1 startup roles") entirely in natural language for highly nuanced targeting.
Explainable AI Matching
No more black boxes. FTJA acts as an intelligent agent, cross-referencing your resume, custom filters, and the JD to provide a transparent, detailed explanation of why a job passed or failed.
Instant Categorization
Every time you click on a job posting, FTJA performs real-time analysis in parallel in the background, rapidly categorizing them into "Apply Now," "Review First," or "Skip."
MVP Built for Fast Learning Cycles
To ensure the best user experience, we built the MVP as a Chrome Extension. It integrates directly into existing job platforms like LinkedIn, allowing users to access features via a side panel without needing to adapt to a new interface.
Upload resume & documents, then set job preferences.
Click through multiple job postings in parallel.
Review transparent match reasons, semantic scores, and actionable feedback directly in the browser.
Finding the Perfect Fit
"Developed during the Perplexity AI Business Fellowship, FTJA goes beyond simple API wrapper chatbots. It tackles a massive inefficiency in the job search funnel—where 95% of listings are irrelevant. By architecting an LLM + vector search pipeline and designing continuous evaluation frameworks, we achieved 95%+ matching accuracy at a highly scalable cost of $0.01 per query."