Skip to main content
DIMAG AI LABS

// Auto-Scientist

The AI lab for
neural systems,
behavior, and discovery.

We build an auto-scientist for neuroscience teams: a platform for standardized data, neuro foundation models, multimodal experiment analysis, and scientific agents that uncover patterns across calcium imaging, neuronal recordings, behavior, and task structure.

Formats
NWB

Standardized metadata and trace provenance for neuroscience work.

Signals
Calcium

Imaging, spikes, traces, and population-level neural codes.

Research stack
AgentsModelsExperiments
Agents

Research agents for hypothesis generation, experiment planning, feature extraction, and condition comparison across large multi-session datasets.

Pre-trained Models

Neuro-specific encoders and task-adaptable model backbones for calcium imaging, behavior, and neural time series analysis.

Experiments

Experiment toolkits that manage trial design, variable alignment, behavioral scoring, and evidence generation from live results.

Data Analysis

Analytical workflows for denoising, segmentation, dimensionality reduction, statistics, and cross-modal interpretation.

// Thesis

A real Auto-Scientist should do the work of a research system, not just a chatbot.

The future of neuroscience is not just better imaging; it is better scientific systems that ingest noisy signals, standardize metadata, and turn experiments into interpretable evidence.
We believe the auto-scientist should function as a collaborative research partner: building datasets, training models, designing experiments, and generating hypotheses with experiment-grade traceability.
That means an operating layer that supports calcium imaging, neuronal activity, behavioral tracking, multimodal recordings, and NWB-compatible pipelines in one unified platform.

// Data standards

Proper formats for real neuroscience pipelines.

NWB 2.x

The standard for experiment metadata, neural recordings, behavioral streams, stimulus annotations, and reproducible processing history across labs and collaborators.

Calcium Imaging

2P, 1P, miniscope, and widefield imaging pipelines with motion correction, ROI detection, deconvolution, signal denoising, and session alignment.

Neuronal Electrophysiology

Spike sorting, LFP analysis, multi-unit recordings, and event-level alignment to trials, timing windows, and sensory conditions.

Behavioral Tracking

Pose estimation, locomotion metrics, gaze tracking, social interaction analysis, and action segmentation from video and sensor streams.

Imaging + Signals

Integrated time-series fusion between neural activity, calcium traces, imaging volumes, and experimental variables for meaningful causal analysis.

Metadata & Provenance

Audit-ready experiment logs, subject metadata, hardware settings, software versions, and pipeline lineage to support reliable science.

// Neuro foundation models

Models built for neural activity, behavior, and imaging.

Foundation model layer

Neuro Foundational Models

Pretrained encoders for neural sequences, imaging time series, and behavior trajectories that can be fine-tuned for new animals, tasks, and experimental paradigms.

Time-series encodersBehavior latent spacesImaging backbones
Foundation model layer

Behavioral AI Agents

Agents that parse behavior, infer action states, detect anomalies, and connect movements to neural activity for high-level scientific hypotheses.

Action detectionPose analysisTrial inference
Foundation model layer

Calcium & Neuronal Imaging

End-to-end workflows for denoising, segmentation, deconvolution, spatial mapping, trial alignment, and population-level communication analysis.

ROI extractionDenoisingPopulation coding
Foundation model layer

Experiment Planning Agents

Planning agents that suggest hypotheses, define conditions, prioritize experimental cohorts, and automate evidence gathering from live analyses.

Hypothesis scoringMulti-arm planningCondition ranking

// What your lab gets

A complete neuroscience research operating layer.

Standardized, searchable neuroscience datasets in NWB and lab-native formats

Calcium imaging and neuronal signal analysis pipelines with QC, segmentation, and deconvolution

Behavioral models that connect motion, pose, and action sequences to neural activity

Pre-trained neuro foundation models that adapt to new experiments, animals, and tasks

Experiment orchestration agents for candidate testing, condition comparison, and insight generation

Statistical summaries, visual reports, and reproducible analysis artifacts for publication workflows

Need an AI platform for your neuroscience lab?

We support labs that need data standardization, multimodal analysis, pre-trained models, and experiment agents built for real scientific discovery across imaging, neuronal activity, and behavior.

Auto-Scientist for neuroscience labs

Neuro foundation models, NWB-compatible data pipelines, calcium imaging workflows, neuronal analysis, behavioral modeling, and experiment agents for scientific discovery.