Team
LeetCoach
Project Concept
Problem
LeetCode’s Easy / Medium / Hard buckets are too coarse; learners waste time bouncing between problems that aren’t actually easier.
Solution — LeetCoach
A headless CLI coach that uses local Zerotrac numeric difficulty to build real practice ladders.
Given a LeetCode URL (and optional WIP code), Gemini will:
- Explain the problem — intuition → steps → complexity.
- Suggest 2–3 strictly easier bridge problems by rating (with guardrails to prevent “accidentally harder” picks).
- Optionally review your code with a minimal counterexample and tiny patch, then let you choose: solve now or try bridges.
Why Local?
Ratings and statements are files on disk — reproducible, private, and fast.
No scraping or production deployment required.
Gemini Integration
- Non-interactive prompts
- File grounding
- (Optional) MCP shim for ratings lookup
Impact
A reusable pattern for local, data-aware agents that make learning workflows measurably better.
Entry
Status: Submitted
Last saved: September 06 at 3:15 PM EDT
Team Roster
Message board not available for this team yet.
Roger Sentongo Team Lead RSVP Approved
Software Engineer at Contrivance Inc
Designed and developed the product.
I was born and raised in Uganda, did all my university education in the US at the University of South Florida and City University of New York. Very interested in multimodal AI systems and their applications.
Multimodal AI and its applications. Working on some software products in this field
www.pulse-nyc.com
Thats my main project at the moment