
Aldeas: BlazeBPF, The AI-Powered Way To Learn eBPF
We strongly believe that “Preparation is what turns potential into performance.” Effective playgrounds should be designed to allow students to learn independently, not just by following instructions. However, if the gap between a student's current skills and those required to solve the problem is too wide, they are likely to stagnate and drop out. We've seen it countless times. That's why we created BlazeBPF, a platform for learning eBPF, a Linux kernel technology for safe, efficient dynamic programming.
eBPF provides a significant competitive advantage for developers working with the Linux kernel. It allows them to extend and customize the operating system's behavior in real time without modifying the kernel source code or restarting the system.
Blazebpf supports your development and learning journey with an innovative AI-powered approach.
- Adaptive difficulty
- Real-time intelligent assistance
- Progressive hint system
- Context-aware guidance
- Dropout rate target: <20%
The AI Companion is an always-available tutor that understands:
- What the student is working on
- Where they're stuck
- Their skill level
- Their learning history
Why this matters
eBPF is a hot topic.
Students will master Linux kernel programming through hands-on challenges, real-time compilation, and a progressive skill tree.
BlazeBPF is a business solution that offers four levels that drive user acquisition and value proposition for individuals, communities, and businesses.
- FREE - $0/forever (Always Free!)
- PRO - $9.99/month or $99/year
- TEAM - $49/user/month
- ENTERPRISE - Custom pricing
How we built this
BlazeBPF is built on AWS with a highly scalable, multi-tier architecture designed to handle growth from hundreds to millions of users. The infrastructure uses managed services for reliability and auto-scaling capabilities for cost-effective growth.
Looking at the project's history, we went through three stages: the first was traditional development with no AI tools; then we used Copilot to improve productivity; and finally, we used Kiro to further streamline it. We realized there's no going back: Kiro marks a new direction in AI-driven software development, supported by the full AWS AI Portfolio and services.
BlazeBPF uses three AI-powered systems working together to bridge the skill gap:
- AI Learning Companion (AWS Bedrock - Claude)
- Intelligent Hint System (Progressive Scaffolding)
- Adaptive Content Curation (AI-Powered)
AI-supported learning.
Context-Aware:
- Knows which lab/challenge the student is on
- Understands the learning objective
- Sees the student's code and errors
- Tracks progress and struggle points
Pedagogically Sound:
- Explains concepts, doesn't just give answers
- Uses Socratic method (asks guiding questions)
- Provides examples at an appropriate level
- Encourages independent thinking
Emotionally Intelligent:
- Detects frustration (time spent, repeated errors)
- Provides encouragement
- Celebrates small wins
- Maintains positive tone
Adapts to Individual:
- Fast learners get minimal hints
- Struggling students get more support
- Everyone learns at their own pace
Key Features
✅ Global CDN - CloudFront for low-latency worldwide access
✅ Auto-Scaling - EC2 Auto Scaling Groups for compute
✅ Serverless Database - DynamoDB for unlimited scale
✅ Isolated Compilation - Secure eBPF compilation in private subnet
✅ Multi-Region Ready - Architecture supports global expansion
✅ AI Agents - as previously mentioned
High-Level Architecture



Multi-Region Architecture (Future)

At this time, we do not use EKS, Cognito, WAF, among other services; we strive to minimize expenses and over-engineering.
Demo
The best way get a real sense of our proposal is through www.blazebfp.com . Just remember that it is not production-ready yet. If you find a bug or have any questions or comments, please let us know.
Screenshots
Here we leave some screenshots for your quick reference.


Building BlazeBPF from concept to production taught us invaluable lessons about full-stack development, cloud architecture, educational technology, and building a sustainable SaaS business.
Disclaimer. We selected a few scenarios because we have many, many more.
🎯 Technical Learnings
eBPF Compilation is Complex - Simplification is Key
Challenge: eBPF requires specific kernel headers, BPF helpers, and complex toolchains that are difficult to replicate in a web environment.
What We Learned:*
- Minimal Injection Strategy: Instead of providing full kernel headers (which would be 100MB+), we created a "minimal injection" system that adds only essential definitions (SEC macro, protocol constants, byte order macros)
- Progressive Enhancement: Start with basics, add features incrementally based on actual user needs
- Guard Macros: Use
#ifndefguards to avoid conflicts when users include their own headers
- Built-in Functions: Leverage compiler built-ins (
__builtin_bswap16) instead of manual implementations for efficiency
Key Insight: "Perfect is the enemy of good." A 67% success rate with minimal injection is better than 0% with no solution. We can improve incrementally.
Authentication & Authorization: Security is Hard
Challenge: Building a secure auth system with JWT, sessions, and role-based access control.
What We Learned:
- JWT for Stateless Auth: Store user ID and role in token, verify on every request
- Refresh Tokens: Short-lived access tokens (15 min) + long-lived refresh tokens (7 days)
- HttpOnly Cookies: Store tokens in HttpOnly cookies to prevent XSS attacks
- CORS Configuration: Strict CORS policies prevent unauthorized access
- Rate Limiting: Prevent brute force attacks with Redis-based rate limiting
- Password Hashing: Use bcrypt with a high cost factor (12+)
AI Integration: Context is Everything
Challenge: Integrating AWS Bedrock (Claude) for intelligent learning assistance.
What We Learned:
- Context Matters: AI needs to know what lab the student is on, what they're struggling with, and their history
- Prompt Engineering: Well-crafted prompts make the difference between helpful and useless responses
- Progressive Hints: Don't give away answers immediately. Guide students through problem-solving
- Fallback Strategies: Have static hints when AI is unavailable or slow
- Cost Management: Cache common queries, limit requests per user
Architecture Evolution:
Phase 1 (MVP):- 1 EC2 instance (backend + compilation)- S3 for static assets- DynamoDB for data
Phase 2 (Current):- 2 EC2 instances (backend public, compilation private)- CloudFront for global CDN- ElastiCache for session management- Multiple S3 buckets for content organization
Phase 3 (Future):- Auto Scaling Groups- Application Load Balancer- Multi-region deploymentKey Insight: "Scale when you need to, not when you think you might." Premature optimization wastes time and money.
Cost Impact: Current infrastructure costs $300-400/month for 100-500 users. Can scale to 1M+ users with the same architecture.
🎓 More than the technical.
- Technical Excellence Matters: But perfect is the enemy of shipped
- User Experience is King: Interactive learning beats passive watching
- Community is Your Moat: Free users become advocates and contributors
- Business Model Matters: Freemium with a clear value proposition works
- Documentation Saves Time: Future you will thank present you
- Automation is Essential: Manual processes don't scale
- Monitoring is Critical: You can't fix what you can't see
- Focus is Hard: Saying no to good ideas is necessary
- Persistence Wins: Keep shipping, keep learning, keep improving
- Learning Never Stops: Every day brings new challenges and lessons
The most important lesson: Building a startup is not about having all the answers. It's about being willing to learn, adapt, and persist through challenges.
Key Insight: "You don't need to be an expert in everything, but you need to know enough to be dangerous."
I hope you enjoy this project; we've been building it with dedication and care.
If it's not too much trouble, we'd appreciate a thumbs up.
And any comments, questions, or suggestions, please leave them in the comments section.
regards,
Alejandro Mercado y Alvaro Soto
Now go build something amazing! 🚀





Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article