TL;DR: Vector search with end-to-end encryption was theoretically impossible - queries need plaintext to compute distances. We solved it. Now we're challenging developers to build fully encrypted AI systems that perform at scale. $10,000 in prizes across two phases. Registration closes November 23.
The Challenge: Build What Security Engineers Say Can't Exist
There's a common fallacy in security: semantic search on encrypted data is impossible.
The math seems straightforward. Vector similarity requires computing distances between embeddings. Encryption scrambles those numbers. Distance calculations on encrypted vectors return garbage.
We figured out how to do it - with encryption-in-use that keeps embeddings encrypted during search while maintaining sub-10-millisecond latency at scale. Meet CyborgDB.
How CyborgDB Works
CyborgDB uses cryptographic indexing to compute approximate nearest neighbors on fully encrypted vectors. Your embeddings never exist in plaintext-reversible formats - not at rest, in transit, in memory, or in logs.
What you get:
- Customer-controlled keys (BYOK/HYOK)
- Drop-in proxy for Postgres/Redis
- No application rewrites required
- Sub-10ms encrypted queries at scale
We've tested this internally. But nobody's built production AI systems with fully encrypted vector search before - because until now, it wasn't possible.
We're opening CyborgDB for 8 weeks across two phases. Build something that shouldn't exist.
Two Phases: Ideas → Working Prototypes
Phase 1: Idea Submission (Due November 23, 2025)
Submit a 2-3 page proposal. We're looking for use cases where encrypted vector search unlocks real business value.
Phase 1 Prizes - $1,500 Total:
- Best Healthcare/HIPAA Use Case: $500
- Best FinTech/Compliance Use Case: $500
- Most Creative Enterprise Use Case: $500
What to include:
- Problem statement: What security blocker prevents this AI application today?
- Proposed solution: How will you use CyborgDB to solve it?
- Business impact: Quantifiable value (revenue unlocked, risk reduced, time saved)
- Technical approach: High-level architecture and integration points
Why Phase 1 matters: Win prizes for your idea AND get feedback from our engineering team before building. Phase 1 winners get priority support during prototyping.
Phase 2: Working Prototype (Due December 20, 2025)
Build a functional demo proving your concept works with real encrypted vector search.
Phase 2 Prizes - $8,500 Total:
- Grand Prize (Best Overall): $3,500
- Best Healthcare Implementation: $1,500
- Best FinTech Implementation: $1,500
- Best Enterprise Implementation: $1,000
- Best Technical Feedback & Documentation: $1,000
All Phase 2 completers receive:
- 3-month CyborgDB Enterprise trial ($2,500 value)
- Direct Slack access to CyborgDB engineering team
- Featured case study opportunity
- Early access to roadmap features
Special Recognition: Technical Feedback Award
$1,000 for Best Technical Feedback & Documentation
Found a performance bottleneck? Documented integration rough edges with LangChain? Discovered edge cases we didn't consider?
What we're looking for:
- Detailed performance insights (latency breakdowns, scaling limits, throughput under load)
- Documentation feedback (what worked, what didn't, workarounds you needed)
- Bug reports or feature gaps that would improve production readiness
Honest technical feedback helps everyone - including future users. We want to know where CyborgDB breaks, not just where it works.
Timeline
Idea Phase: October 28 - November 23, 2025
Hackathon Phase: November 26 - December 20, 2025
Why You Should Do This
You're building AI in healthcare, fintech, or enterprise - and security is blocking your roadmap.
Prototype the secure version you actually need, with support from engineers who understand the problem.
You care about hard technical problems.
Encrypted similarity search at scale is genuinely novel - there are maybe 5 people in the world who've built this. You could be one of them.
You want early adopter credibility.
When encrypted AI becomes table stakes (regulatory pressure will force this), you'll be the person who built real systems with it before anyone else.
You need a standout portfolio project.
"Built encrypted RAG system for healthcare AI" beats "built another todo app" every time.
You're looking for your next thing.
Direct access to our engineering team, showcase at demo day, and portfolio pieces demonstrating real technical depth open doors.
You Don't Need To Be An Expert
- You don't need cryptography expertise - CyborgDB handles that
- You don't need a team - solo projects are welcome (but teams of 2-4 are ideal)
- You don't need production data - synthetic datasets from HuggingFace and Kaggle work fine
You need to understand vector search, embeddings, and how to build AI applications.
Get Started
Register: hackerearth.com/challenges/hackathon/cyborg
Read the docs: docs.cyborg.co
Questions? Email hello@cyborg.co
Registration closes November 23, 2025. Don't wait - the sooner you start, the more time you have to build something impossible.
P.S. For Design Partners
If you're working on something in healthcare, fintech, or enterprise that needs encrypted vector search - but can't commit to the hackathon timeline - reach out anyway.
We're looking for design partners. This might be exactly what you need.
Contact: hello@cyborg.co
