What is a Large Behavioural-omics Model (LBM)?

A Large Behavioural-omics Model (LBM) is a foundational AI system that models who you are — not just what you do. Unlike LLMs that model language distributions, LBM models behavioural state trajectories, creating a living digital twin that evolves with every interaction.

How is LBM different from traditional recommendation systems?

Traditional AI predicts your next word or click but cannot model WHY you behave the way you do. LBM uses causal reasoning and continuous learning to understand, predict, and adapt to individual human behaviour through behavioural digital twins.

Is my data private with Dots-In?

Yes. Dots-In uses privacy-preserving techniques including homomorphic encryption, federated learning, and differential privacy. Your raw data never leaves your device — only encrypted gradients are processed.

Dots-In Research Labs

Modelling the
human,
not the text.

We build the Large Behavioural Model (LBM) — the world's only foundational AI architecture that treats human behaviour as a first-class computational primitive.

Explore our research →Join the community

Mission & Vision

Why we exist.

Language has LLMs. Vision has ViTs. The physical world has JEPA. Human behaviour has no foundation model. We're here to change that.

Mission

Build the world's first foundational AI that understands human behaviour — not through clicks and keystrokes, but through the causal, temporal, multi-dimensional structure of how humans actually function.

Vision

A world where every AI system — from healthcare to education, from productivity to mental health — is powered by a shared behavioural intelligence layer that truly understands the human it serves.

Modality	Input	Architecture	Status
Language	Tokens	GPTClaudeLLaMA	✓ Solved
Vision	Pixels	ViTsCLIPDINOv2	✓ Solved
Physical World	Actions / Physics	JEPAGenie	◐ Emerging
Audio	Spectrograms	WhisperAudioPaLM	✓ Solved
Human Behaviour	Behavioural State Vectors	LBM	◉ We're building it

Ongoing Research

What we're working on.

Six core research streams, each addressing an unsolved problem in modelling human behaviour at scale.

Behavioural State Vectors

The 240-dimensional continuous representation of a human's current state — cognitive, emotional, biological, motivational — updating in real-time.

Active

Causal Behavioural Graphs

Directed causal graphs capturing why behaviour happens — 230M+ validated edges across 8 discovery methods.

Active

Episodic Memory Architecture

Multi-scale memory compressing 15-minute episodes and 15-year developmental arcs into a unified retrieval system.

In Progress

Self-Supervised Learning

7 self-supervised objectives for training on behavioural data without ground truth — next-token prediction for behaviour.

Active

Safe Reinforcement Learning

Mathematical safety guarantees for behavioural interventions. When should a model nudge, warn, or stay silent?

In Progress

Population Intelligence

Scaling individual models to population-level insights without Simpson's paradox — from n=1 to n=10M.

Forming

Architecture

How LBM works.

A five-layer architecture transforming raw signals into causal understanding and safe intervention.

LBM Architecture Stack

Each layer builds on the one below — from raw signal to actionable understanding.

Ingestion

Digital activitySleep patternsPhysiologyEnvironmentSocial signals

Episodes

Micro (15-30m)Macro (days-weeks)Context windowsState inference

BSV Engine

240-dim vectorReal-time updatesLatent spaceTemporal dynamics

Causal Graph

8 discovery methods230M+ edgesRoot-cause inferenceCounterfactuals

Intervention

Safe RL policyNudge / Warn / SilenceSafety boundsLong-horizon RL

Built by the Dots-In founding research team.

The Team

Who's building this.

A small, focused founding team with deep expertise across AI, behavioural science, genomics, and quantitative finance.

Suraj Biswas

Founder & CEO

Behavioural AI systems

Saurav Gupta

Co-founder & CTO

IIT Kanpur · AI systems engineering

Aruna Dey

Co-founder & CPO

Genomics

Events

AI's Missing Primitive.

A weekly invite-only roundtable in Bangalore. Every Saturday, a different facet of behavioural AI — debated, challenged, built in the open.

Mar

The Behavioural Primitive

Session completed

Completed

The Lie Detector Test That Doesn't Lie

How AI can learn you without asking you anything

Upcoming

May

The Memory Problem

How should AI remember you — 15 minutes or 15 years?

Upcoming

Jun

When AI Knows You Better Than You Do

The ethics of behavioural prediction

Upcoming

Jul

One in a Billion, But Still Predictable

Why individual behaviour scales to population intelligence

Upcoming

Aug

Behavioural Intelligence at Scale

Roundtable session

Upcoming

Publications & Open Source

Our work.

Papers, proposals, and open-source artifacts from the community.

Paper

Large Behavioural-Omics Model (LBOM) — Technical Dossier

Foundational architecture notes

Paper

LBM Technical White Paper — Architecture, Formalisms & Validation

Pre-publication · Access under NDA via community onboarding

Proposal

IndiaAI Mission — National Foundational AI Model Proposal

Submitted January 2026 · Behavioural intelligence infrastructure

Open Source

Behavioural Data Standards v0.1

Community Build Workshop · June 2026

Open Source

CBG-Viz — Causal Behavioural Graph Visualisation

Community Build Workshop · June 2026

Open Source

BehavioralBench — Evaluation Benchmark for Behavioural AI

Community Build Workshop · June 2026

The room is open.

We're building the first Behavioural Foundation Model from India — not as a company project, but as a national scientific effort.

Join the community →

Modelling thehuman,not the text.

The room is open.

Modelling the
human,
not the text.