Career PathFinder: Explainable AI Career Roadmaps
Mayank Tiwari
Adaptive roadmap: every milestone is ordered by a prerequisite graph and comes with a glass-box trace of why it's there.
The 53-skill prerequisite DAG, sequenced with Kahn's topological sort: 0.00% prerequisite violations across 1,000 simulated learners.
Bayesian Knowledge Tracing diagnostic: measures the probability you've mastered each skill instead of trusting a 1-5 self-rating.
Top 25 team at HCLTech AMPlified: The AI Challenge, Season 1 (Round 2), presented to the HCLTech panel.
Ask a chatbot how to become a data analyst and it invents courses, puts deep learning before Python, and takes your word that you "know SQL". Career PathFinder fixes all three with maths, not prompts.
Measures real mastery. Bayesian Knowledge Tracing (the Corbett-Anderson model) estimates the probability you've mastered each skill from a 139-item diagnostic, instead of a 1-5 self-rating.
Never gets the order wrong. A 53-skill prerequisite graph sequenced with Kahn's topological sort: 0.00% prerequisite violations across 1,000 simulated learners.
Prioritises what matters. Skill gap = O*NET importance x (1 - mastery), so you study what the target role weighs most. Courses come from 105 real, curated options ranked by semantic embeddings.
Explains every step. Each milestone carries a glass-box trace. Gemini is fenced to parsing goals and phrasing explanations; it never picks the curriculum.
FastAPI · scikit-learn · sentence embeddings · Gemini · 54 tests, OWASP-hardened, CI on every push.
Like this project
Posted Oct 16, 2026
Top 25, HCLTech AMPlified S1. Bayesian Knowledge Tracing + a 53-skill prerequisite DAG turn a career goal into a roadmap with 0% prerequisite violations.
Likes
2
Views
1
Clients
HCLTech
