HelixQA
Test your AI before your users do.
- Role
- Product and frontend engineer
- Type
- Team
- Year
- August 2026 to Present
- Status
- In development
- Category
- AI and Tools
- Main stack
- Next.js, TypeScript +6
Overview
HelixQA is a quality platform for teams shipping AI. It checks products for correctness, groundedness, instruction following, safety and regressions, so failures are caught before users meet them.
Teams write test cases by hand or generate candidates with AI and approve them before saving. Suites run on demand, on a schedule or inside CI, and each run produces scores per quality dimension with failures and diffs. A REST API, webhooks and SDKs fit it into an existing pipeline.
I build the product surface at Nexusbert: the marketing site on helixqa.tech and the frontend that sits on top of the HelixQA backend. The backend is provider agnostic and uses adapters, so customers are not tied to one model vendor.
Highlights
- •Use cases for support AI, RAG assistants, internal copilots and agent workflows.
- •Deterministic scoring across testing, security, performance, quality and stability.
- •Regression detection by comparing versions before a release.
- •Workspace scoped API keys, access controls and data retention settings.
Built with
- Next.js
- TypeScript
- Tailwind CSS
- Cloudflare
- Python
- FastAPI
- PostgreSQL
- Docker
