Coding Problem Recommender
Semantic search system that maps software engineering job descriptions to relevant LeetCode problems using embeddings and clustering.
Project information
- Category: Semantic Search / Interview Prep
- Dates: Jan 2025 - Apr 2025
- Tech: Python, Sentence Transformers, Scikit-Learn, NumPy, FastAPI, Streamlit, Azure ML, Grok API
- GitHub: Interview-prep-tool-AI-ML
Project Details
Overview
The Intelligent Coding Problem Recommendation System maps job descriptions to relevant LeetCode problems so candidates can focus interview prep around role-specific requirements instead of generic topic lists.
Problem
Software engineering job descriptions vary by stack, seniority, and domain. A strong prep workflow should understand those requirements semantically and recommend practice problems with topic diversity.
Engineering
- Processed and embedded 1,571 LeetCode problems with Sentence Transformers for semantic retrieval.
- Implemented cosine similarity ranking and K-Means clustering to improve relevance and avoid repetitive recommendations.
- Built a FastAPI backend and Streamlit interface with LLM-generated explanations for why each problem matches the input role.
- Used Azure ML for deployment experimentation and model workflow management.
Impact
Delivered an interactive recommendation workflow with relevance scoring, cluster-aware diversity, and clear explanations that connect job requirements to coding practice.