AI research and products for systems people can rely on

Obvix Labs builds AI systems that stay honest, grounded, and controllable. Then we publish what we learn.

Amy AI

An AI companion that pushes back when it matters, tracks emotional patterns, and never tells you what you want to hear. Built with four-layer detection for cognitive distortions, crisis protocols, and dependency warnings.

Learn more

Crisis detection with 96%+ recall

Agreement ratio: 15-25%

Anti-sycophancy by design

What we're building

Products

View all

Testing

Obvix Proxy

Real-time LLM safety shield

A drop-in proxy that intercepts every exchange between users and language models, making context-aware decisions to block or edit harmful content without retraining the underlying model.

Context-aware filtering + rule checks
~58 ms latency on consumer GPU
Works with any LLM provider

Prototype

Obvix Lake

GraphRAG customer support platform

An AI-native customer lake that fuses knowledge graphs, RAG retrieval, sentiment sensing, and ticket learning loops for grounded support automation.

Knowledge graphs + RAG retrieval
Sentiment & intent sensing
FAQ auto-learning loop

Research

Latest from the lab

All articles

AI Safety2026-03-18

The Behavioral Ratchet: What 80,433 Trials Reveal About LLM Sycophancy

We ran 80,433 trials across six models to answer a specific question: does sycophancy increase as the context window fills up? The answer surprised us. Context length matters, but only for small models. The real driver is conversational pattern — and it is reversible.

Karan Prasad

7 min read

Principles

What guides our work

Safety first

Every system starts with the question: what could go wrong? We design guardrails before features.

Radical transparency

We publish our methods, share our playbooks, and invite scrutiny. Safe AI needs a bigger commons.

Human-centered

AI should augment human judgment, not replace it. We build escalation paths and benevolent friction into every workflow.

Collaborative rigor

We co-design with policymakers, clinicians, and civic groups so solutions reflect the needs of those they affect.

Join us

We're looking for people who care about getting AI right

Researchers, engineers, and collaborators who want to build AI systems the world can actually trust.

Meet the team Get in touch