README.md

# SIMAM AI LAB

### Applied Intelligence for Industry, Infrastructure & Public Systems

**AI Agents · Spatial Intelligence · Local AI · Vision · Digital Twins · XR**

Simam AI Lab is the applied AI research division of **Simam Digital Ltd**, a UK digital product and engineering studio based in Wakefield.

We research, prototype and publish practical AI systems that bridge the gap between emerging models and real-world deployment.

Our work focuses particularly on AI that can **understand data, reason about physical environments, use tools and assist people inside complex operational workflows.**

---

## 🧪 What We're Exploring

### Agentic Infrastructure

Reusable AI agents and orchestration systems for engineering, construction, infrastructure, public services and enterprise workflows.

### Spatial Intelligence

AI systems capable of reasoning across **Gaussian Splats, GIS, BIM, maps, Digital Twins, imagery and real-world environments**.

### Autonomous Workflows

Multi-stage systems combining models, tools, agents, APIs and human approval to produce useful operational outputs.

### Local Intelligence

Private and on-premise AI using open models for organisations that need greater control over their data, infrastructure and inference.

### Human + AI Interfaces

Exploring how **voice, vision, XR, spatial computing and intelligent interfaces** change the way people interact with AI.

---

# 🤗 AI Model Playground

We use Hugging Face to make emerging AI research easier to explore.

Our Spaces include simple interfaces for experimenting with:

* Large Language Models
* Vision Language Models
* OCR & document intelligence
* Image understanding
* Object detection & segmentation
* Embedding models
* Local and open-source LLMs
* AI agents
* Retrieval and RAG
* Generative image & video models
* Spatial AI experiments

The goal is simple:

> **Make powerful AI research accessible without requiring users to understand Python, CUDA, inference servers or model deployment.**

---

# 🧠 Simam Intelligence Framework

Our applied research follows a simple operating loop:

**OBSERVE → UNDERSTAND → REASON → ACT → LEARN → IMPROVE**

We combine four practical layers:

**01 — Spatial & Data Signals**
3DGS · BIM · GIS · imagery · documents · sensors · APIs

**02 — Intelligence**
Vision models · LLMs · multimodal models · agents

**03 — Workflows**
Tools · orchestration · human approval · evaluation

**04 — Deployment**
Cloud · edge · local · on-premise

Models are selected around the problem — not the other way around.

---

# 🔬 Current Research

Some of our active research areas include:

**AI Construction Assistant**
Spatial agents navigating Gaussian Splat construction environments for inspection, safety and progress analysis.

**Local AI for SMEs**
Privacy-preserving local language and vision models for organisations working with sensitive information.

**XR AI Training**
Vision, voice and spatial agents operating inside immersive industrial training environments.

**Infrastructure Intelligence**
Combining mapping, Digital Twins, operational data and AI agents to create queryable infrastructure systems.

**Agentic Workflows**
Research into visual, secure and human-supervised agent orchestration for non-technical users.

---

# 🧩 Simam Agent Studio

We are developing a visual Applied Intelligence platform for creating AI agents and multi-stage workflows without requiring traditional AI development.

Think:

`INPUT → MODEL → REASON → TOOL → APPROVAL → ACTION`

Our research explores:

* Visual agent orchestration
* Model-independent workflows
* Bring-your-own-model / bring-your-own-key
* MCP and external tools
* Human approval gates
* Agent evaluation
* Local model execution
* Multi-agent systems
* Industry-specific Agent Blueprints

---

# 📊 Models & Benchmarks

Not every AI model is right for every problem.

We experiment with open models and publish practical observations around:

* reasoning quality
* vision performance
* document understanding
* latency
* hardware requirements
* inference cost
* privacy
* local deployment
* agent/tool performance
* industry suitability

Where useful, experiments will be released as **Spaces, datasets, benchmarks or Research Notes**.

---

# 🛰 Applied Spatial Intelligence

A major focus of the Lab is exploring what happens when AI begins to understand the **physical world**, not only text.

We research combinations of:

`AI + GIS + 3DGS + BIM + Digital Twins + Computer Vision + XR`

Our long-term goal is to create intelligent systems capable of understanding complex physical environments and helping people **inspect, navigate, simulate, query and operate them using natural interfaces.**

---

# 🤝 Research & Collaboration

We are interested in collaborating with:

* AI researchers
* Open-source developers
* Universities
* SMEs
* Infrastructure organisations
* Engineering teams
* Public-sector organisations
* AI model developers
* Research funders
* Technology partners

We are particularly interested in applied research where an emerging AI capability can be tested against a **real operational problem**.

---

## Build With Us

Have a model, dataset, research idea or real-world problem worth exploring?

We're open to:

**Research collaborations · Model evaluations · Dataset projects · Industry pilots · Agent experiments · Spatial AI research · Open-source projects**

🌐 **Simam AI Lab:** lab.simamdigital.com
🏢 **Simam Digital:** simamdigital.com
💻 **GitHub:** github.com/Simam-Digital-Ltd
📍 **Wakefield, United Kingdom**

---

### Research → Prototype → Evidence → Product

**Simam AI Lab**
*Applied intelligence for the physical world.*