AI strategy and custom software,
built for your business.

Bring us the problem you want to solve. We help shape your AI strategy and turn ambitious ideas into intelligent tools, connected workflows, and software that works in the real world.

Start a conversation Explore our work

Research depth. Engineering expertise. Built around you.

From a first prototype to an integrated system.

Custom developmentApplied researchEnd-to-end delivery

Selected work / possibilities ahead

Proven capabilities.
New possibilities for you.

Built for clients. Tailored for you. These examples demonstrate our capabilities. We build custom solutions for each client; the systems shown are not offered as licensed products.

AI tutoring & training

01

We developed a curriculum-aligned AI tutor that helps learners work through problems with step-by-step guidance, real-time graphs, and text, audio, and avatar-based video.

Multimodal AITool callingGuardrails

What we can build for you

Learning assistants, employee training, customer onboarding, or expert guidance grounded in your organization’s content and methods.

Inside the work

The tutoring system combines instructional design with AI orchestration, graphing tools, safety checks, and experimentation. A related teacher interface organizes sessions, highlights flagged activity, and links to transcripts for human review.

We can bring the same approach to domain-specific coaching and guided workflows, with oversight and integrations suited to your users.

Related research: multimodal AI tutoring ↗

AI content authoring

02

We built an authoring workflow that generates structured tutoring content, lets authors refine it, checks consistency, and connects approved material to existing content systems.

Batch generationHuman reviewAPI integrations

What we can build for you

Editorial copilots, knowledge-base publishing, training content pipelines, and document workflows that keep your experts in control.

Inside the work

The workspace supports item-by-item and batch generation, editable goals and hints, consistency checks, revision tracking, and a status board for assignments and quality review. It integrates content sources, shared files, and downstream publishing through APIs.

For your organization, we can connect generation to your own templates, approval stages, and publishing destinations.

Workplace knowledge assistants

03

We developed Slack assistants for employee policies, product knowledge, curriculum resources, and technical guidance, including document retrieval and source-linked answers.

Knowledge retrievalSlack assistantsEnterprise integration

What we can build for you

Internal knowledge assistants, sales enablement, support copilots, and technical help connected to your documents and workplace tools.

Inside the work

Different assistants serve different teams: answering policy questions from supplied documents, explaining product information, retrieving relevant curriculum material, and helping developers navigate AI APIs. Retrieval-based implementations use semantic search to ground answers in internal resources.

We can design the knowledge pipeline, conversation experience, access controls, and maintenance workflow around your organization.

Analytics & AI platforms

04

We developed models that predict outcomes from activity data and integrated Langfuse into a client’s AWS environment to track AI interactions and automate conversation scoring.

Predictive modelingAI monitoringPlatform integration

What we can build for you

Forecasting tools and connected monitoring platforms that help you spot trends, investigate AI failures, and assess quality as usage grows—within your existing infrastructure.

Inside the work

Platform integration: We deployed and configured Langfuse, an existing AI observability platform, across client AWS environments. The integration connects cloud databases, storage, and secure credential management, with repeatable deployments using Terraform.

We also developed an extensible evaluator that groups AI interactions into sessions and writes scores back to the platform. One implemented check surfaces toxicity flags from the tutoring application for session-level review.

Predictive analytics: Our modeling work infers reading ability from learning activity and explores how short-term interaction data anticipates longer-term outcomes, with validation across populations.

We can combine custom analysis with established platforms, adapting deployment, integrations, and quality checks to your business.

Related research: predictive model generalization ↗

AI testing & evaluation

05

We evaluated AI tutor behavior across foundation models and user identities, tested tool-generated outputs, and designed experiments to understand how systems perform in practice.

Model selectionBias auditsExperimentation

What we can build for you

Evaluation suites for your use cases, model comparisons, guardrails, regression checks, and experiments that help you decide what to launch and improve.

Inside the work

Our methods include simulated conversations, human ratings, subgroup analysis, and randomized field trials. We use these tools to identify inconsistencies, compare model behavior, and connect technical choices to user experience.

For your team, we can develop representative test scenarios and a repeatable evaluation process that keeps pace with changes to prompts, models, and integrations.

Related research: foundation-model bias evaluation ↗

Content personalization

06

We applied language models to rewrite math content for clarity and explored personalized problem generation that incorporates learners’ interests while preserving learning objectives.

Content adaptationPersonalizationConstrained generation

What we can build for you

Audience-specific communications, accessible explanations, personalized training, and content experiences that reflect your users’ needs and interests.

Inside the work

The work combines prompting strategies, readability analysis, and constraints that preserve the underlying task. A related personalization tool lets students co-author a problem’s context through guided conversation.

We can adapt the same pattern to your domain: define what must stay accurate, what can change, and how people should review or personalize the result.

Related research: LLM content rewriting ↗
Related research: user-directed personalization ↗

How we can help

The expertise to take
your idea further.

Strategy & team enablement

AI roadmaps, investment priorities, and practical workshops that help leaders make decisions and teams put AI to work.

AI & software development

Custom assistants, applications, and automated workflows connected to your data, tools, and cloud infrastructure.

Analytics & AI platforms

Predictive models, decision support, and integrated monitoring platforms that turn operational data into useful signals.

Evaluation & technical review

AI quality and bias testing, model comparisons, and technical due diligence to guide engineering and investment decisions.

Research behind the work

Ideas tested.
Expertise you can build on.

Multimodal learning / PaperLiveHint AI: Curriculum-Specific Multimodal AI Math TutorInstructional design, guardrails, and interactive AI guidance.
Agentic systems / PaperAn agentic framework for real-time pedagogical plot generationConnecting conversational AI to real-time graphing tools.
Predictive analytics / PaperGeneralizing Predictive Models of Reading Ability in Adaptive Mathematics SoftwareBehavioral signals, model generalization, and validation.
AI evaluation / PaperEvaluating an AI Tutor for Bias Across Different Foundation ModelsComparing response behavior across models and user identities.
Content adaptation / PaperRewriting Math Word Problems with Large Language ModelsPrompting strategies and readability analysis for constrained rewriting.
Personalization / PaperUsing Generative AI to Foster Student Sense of Belonging in MathematicsGiving users a role in shaping AI-generated content.

About Extragalactic

Research-minded.
Built for the real world.

Founded by Dr. Husni Almoubayyed

Extragalactic Technologies brings together AI strategy, research, and software engineering. We partner with organizations to identify the right opportunities and turn complex challenges into custom systems their people can use.

Founder Dr. Husni Almoubayyed holds a PhD from Carnegie Mellon University and brings experience leading AI development, applied machine learning, and scientific research. That combination shapes our approach: understand the problem deeply, build thoughtfully, and make the result useful.

Our work in education is one expression of that expertise. The same capabilities can power new experiences and better workflows across industries.

Your next project

What could we
build together?

Tell us what you’re trying to achieve, what’s getting in the way, or the idea you’d like to explore. We’ll help you find a practical path forward.

A solution shaped around you.

Start with an idea, an existing system, or a question. We can help you scope the opportunity, develop a prototype, or deliver a custom implementation.