AWS Builder Center
Weekend Showcase Challenge: AI DSA Problem Recommender

Weekend Showcase Challenge: AI DSA Problem Recommender

AI DSA Problem Recommender is a serverless web application that helps students prepare for coding interviews smarter, not harder. It uses Amazon Bedrock's Nova Lite model to analyze your solved problems and weak topics, then recommends the most impactful DSA problems to focus on next. Features include company-wise filtering (Juspay, Amazon, Google), progress tracking, and topic-based difficulty categorization all deployed entirely on AWS Free Tier services.

Student Fellow @Databricks | AWS Certified (x5) | Cloud - AI - DevOps - SDE | Microsoft (x5) |

Vision & What It Does:

Preparing for coding interviews is overwhelming. With hundreds of problems scattered across platforms, students often waste time solving random questions instead of focusing on their weak areas. AI DSA Problem Recommender solves this by using artificial intelligence to analyze your progress and recommend the most impactful problems to solve next.
The app works like a personal DSA coach. You tell it your weak topics maybe Graphs or Dynamic Programming and as you solve problems and track your progress, the AI engine (powered by Amazon Bedrock's Nova Lite model) identifies patterns in your preparation and suggests problems that will strengthen your weakest areas first. It also lets you filter problems by company (Juspay, Amazon, Google, Microsoft) so you can target your preparation for specific interviews.
From a user perspective, the workflow is simple:
  1. Open the app and set your weak topics
  1. Click "Get AI Recommendations" to receive personalized suggestions with explanations
  1. Solve problems and mark them as completed
  1. Watch your progress grow on the dashboard

How I Built It:

The idea came from my own coding interview preparation journey. I was preparing for Juspay's SDE role and realized I was spending more time deciding what to solve than actually solving problems. I wanted an AI that could make that decision for me.
Key Decisions:
  • I chose a serverless architecture to keep costs at zero (AWS Free Tier) and eliminate server management.
  • Amazon Bedrock with Nova Lite was selected for AI recommendations because it offers strong reasoning capabilities at minimal cost, and it's fully managed no ML infrastructure to maintain.
  • The frontend is a single static HTML file hosted on S3, making deployment trivial and the app lightning fast to load.
  • DynamoDB was chosen for progress tracking because of its pay-per-request billing model perfect for a personal tool with unpredictable usage patterns.

Challenges:

  • The biggest challenge was getting Bedrock to return consistently structured JSON responses. I solved this with careful prompt engineering, including explicit format instructions and a fallback mechanism if parsing fails.
  • CORS configuration between the S3-hosted frontend and API Gateway required careful header setup in the Lambda response.
  • Designing a problem bank that covers enough topics while staying manageable for a weekend project — I curated 20 high-quality problems across 7 topics, each tagged with difficulty and company.

AWS Services Used & Architecture Overview:

The application uses a fully serverless architecture deployed entirely on AWS:
User → S3 Static Website → API Gateway (HTTP) → Lambda (Python 3.12)
↓
┌───────────────┐
│ Amazon Bedrock │ (AI Recommendations)
│ (Nova Lite) │
└───────────────┘
↓
┌───────────────┐
│ DynamoDB │ (User Progress)
└───────────────┘
↓
┌───────────────┐
│ S3 Bucket │ (Problem Bank JSON)
└───────────────┘
AWS Service
Purpose
Amazon S3
Hosts the frontend as a static website AND stores the problem bank (JSON)
Amazon API Gateway
HTTP API that routes requests to Lambda
AWS Lambda
Python 3.12 function handling all backend logic — recommendations, progress tracking, filtering
Amazon Bedrock (Nova Lite)
Generates personalized problem recommendations based on user progress and weak areas
Amazon DynamoDB
Stores user progress — solved problems, weak topics, timestamps
All services run within the AWS Free Tier, making this a zero-cost solution for students.

What I Learned Across the Summer:

This summer's AWS Builder challenges taught me three major lessons:
  1. Serverless is incredibly powerful for solo builders. My first challenge (Build an Always-On Agent) taught me that Lambda + EventBridge can replace an entire backend server. This weekend I applied that same philosophy the entire app runs without a single server to manage.
  1. AI integration is simpler than expected. Using Amazon Bedrock's API is as straightforward as making an HTTP call. The hard part isn't the integration it's the prompt engineering. Getting consistent, parseable responses from the AI model required iteration and fallback strategies.
  1. Ship fast, iterate later. Across all three summer challenges, the biggest lesson was that a working MVP deployed in a weekend is more valuable than a perfect app that never ships. This recommender started as 20 problems and a basic UI but it works, it's live, and it solves a real problem.
  1. AWS Free Tier is underrated. As a student, cost is always a concern. But the combination of Lambda, DynamoDB (on-demand), S3, and Bedrock's free tier credits means I can build and deploy production-quality AI apps at zero cost.
Main Page

All Problems

Set Weak Topics

Progress


Link to App & Repo:

Builder Shoutout:

Shoutout to Lewis Sawe whose "Museum That Grows" project from last week's challenge inspired me to think about how AI can create personalized, evolving experiences. His concept of an agent that builds upon previous context directly influenced how my recommender learns from your progress over time.


Built with ❤️ during the AWS Summer Build Series 2026
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article