Independent AI research & engineering

Research
Beyond
Assumptions.

We investigate how intelligent systems reason, learn, and behave—and build the tools to find out.

Explore our research See what we build
Shinraya Tech Labs original emblem: an ivory S, torii gate, mountain, and vermilion sun
SHINRAYA TECH LABSInquiry without shortcuts.
Our fields of inquiryReasoning & learningQuantum computationScientific simulationReliable AI systems

01 / Research

Questions first.
Evidence always.

From imperfect rewards to geometric retrieval, our work starts with a testable question. The outcome stays open.

All research
01 / ResearchPilot & protocol

Reasoning & reinforcement learning

Robust Reasoning RL

Studying correctness and reward overfitting when verifiers are noisy, incomplete, biased, or inconsistent.

Explore the research
02 / ResearchSimulator findings

Quantum & classical computation

Quantum Advantage Simulator

A shared experimental framework for testing quantum methods against classical references, without assuming an advantage.

Explore the research
03 / ResearchNegative result

Retrieval & geometric learning

ATLAS

Testing geometry, graph structure, and persistent topology for multi-hop retrieval—and retaining the negative result.

Explore the research
04 / ResearchImplementation notes

Computational neuroscience

NeuroSim

An implementation platform connecting biophysical neuron models, network simulation, and scientific visualization.

Explore the research

02 / Engineering

Ideas need
working systems.

Instrumentation, developer tools, and experimental runtimes. Engineering that makes the next question possible.

Explore engineering
01

AgentWatch

Runtime evidence, provenance, comparison, and replay for AI-native systems.

PythonAgentWatch on GitHub
02

RepoPilot

A repository agent that edits code, runs tests, and iterates on execution feedback.

PythonRepoPilot on GitHub
03

VEX

Content-hashed incremental builds with task skipping and a shared remote cache.

RustVEX on GitHub

03 / Our approach

An assumption is
a place to begin.
Not a place to stop.

Meet the lab's philosophy
01

Make it testable.

Turn an interesting idea into a question with controls, baselines, and a way to be wrong.

02

Keep the whole result.

Negative outcomes and unfinished studies belong alongside positive findings.

03

Leave a reproducible trail.

Share the code, methods, configurations, and boundaries that make evidence useful.

Built in the open

Good questions deserve
more perspectives.

Reproduce an experiment. Challenge an assumption. Build something useful with us.

Contribute to the work