Skip to content

AI. SOLUTIONS

An AI solutions startup that builds practical AI products, apps and websites for businesses and everyday users.

See our workStart a project

Computer vision, Forecasting & time series, LLM assistants & RAG, Full-stack AI products, Research & evaluation

An AI financial council in beta, healthcare AI published in a peer-reviewed journal, a medical imaging model, an app people use every day, and two products on the way.

BetaBusiness finance

Majlis

An AI financial council for Egyptian founders.

Upload your bank statements, invoices and contracts, then ask a money question in Egyptian Arabic or English. Up to eight AI advisors debate it, from an Analyst and a Risk Officer to a Skeptic and an Auditor, and the Chair gives you one clear verdict with next steps.

AI advisors debate every money decision
8
AI advisors debate every money decision
clear verdict, with next steps
1
clear verdict, with next steps
numbers guessed: tested code does the maths
0
numbers guessed: tested code does the maths
  • Multi-agent AI
  • Calculation engine
  • Egyptian Arabic RTL
  • InstaPay
Try Majlis

In beta. Business finance guidance, not investment, legal or tax advice.

Published researchHealthcare AI

DiaDose

From a photo of a meal to a safe insulin dose.

An end-to-end diabetes platform. It recognises food from a photo, estimates its carbohydrates, forecasts glucose 24 hours ahead and recommends insulin following ADA guidelines, with an AI assistant for nutrition questions.

food recognition accuracy across 101 dishes
86.6%
food recognition accuracy across 101 dishes
average error estimating carbs from a photo
5.2 g
average error estimating carbs from a photo
R² forecasting glucose with a CNN-LSTM
0.986
R² forecasting glucose with a CNN-LSTM
  • EfficientNetV2-S
  • CNN-LSTM
  • LLaMA 3.3 70B
  • FastAPI
Read the paper in Informatica

Research platform — not a certified medical device.

ResearchMedical imaging

Bone Guard

Reading X-rays for osteosarcoma and nine other tumor types.

A deep learning system trained on 3,746 bone X-rays. It detects osteosarcoma, classifies ten tumor types, combines the image with clinical data, outlines the tumor and writes a PDF report for review.

AUC detecting osteosarcoma
0.99
AUC detecting osteosarcoma
accuracy across 10 tumor types
92.8%
accuracy across 10 tumor types
correct type within the top 3 predictions
97.2%
correct type within the top 3 predictions
  • ConvNeXt-Tiny
  • EfficientNet
  • U-Net
  • Grad-CAM

Research prototype — results support specialist review, not diagnosis.

LivePersonal finance

Saving Goose

Your money, guarded.

A free money tracker with a goose in glasses that guards your budget and roasts you when you overspend. It installs from the browser, works offline, and keeps every number on the user's own phone.

to log a payment
1 tap
to log a payment
languages, with full right-to-left Egyptian Arabic
2
languages, with full right-to-left Egyptian Arabic
accounts, servers or ads
0
accounts, servers or ads
  • Progressive web app
  • Offline-first
  • Vercel
Open the app
In developmentHome services marketplace

Osta

Verified craftsmen, competing quotes.

A marketplace where Egyptian homeowners post a repair job and verified craftsmen bid on it. The craftsman app is Arabic-first and built for people who'd rather speak than type.

  • Open bidding instead of fixed prices
  • Voice-note quotes and read-aloud job descriptions
  • A one-time arrival passcode proves who is at the door
  • Identity verification plus in-person vetting
  • Next.js
  • Supabase
  • Progressive web app
In developmentConsumer AI

U.ai

An AI that notices when something matters.

A conversational AI that remembers you across conversations and checks in on its own after something big happens.

  • Memory you can see and edit
  • Proactive check-ins
  • Encryption by default
  • LLMs
  • Memory graph

We take a model from notebook to production, and we’re honest about what it can and can’t do along the way.

  • Recognition, detection and segmentation models — from identifying a meal in a photo to outlining a tumor on an X-ray.

    • Image classification
    • Segmentation
    • Explainable heatmaps
    • Transfer learning

From the first conversation to marketing and running the product. We only move to the next stage when the last one has proven itself.

  1. 1

    Understand

    We learn the problem, look at the data you have, and agree on the number that will prove the work succeeded.

  2. 2

    Prototype

    A working model on real data, benchmarked against simple baselines, so you see results before committing further.

  3. 3

    Build

    We wrap the model in a product: an API, a dashboard or an app, with security and testing built in from the start.

  4. 4

    Launch

    We deploy it and measure how it performs with real users, so every next decision is based on real results.

  5. 5

    Market

    Once the solution works, we market it: positioning, launch content and campaigns that reach the people it was built for.

  6. 6

    Operate & improve

    We also handle the operations: support, monitoring and updates, while we keep improving the product as new data arrives.

Tell us about it. We'll reply with an honest view of whether AI is the right tool, and what it would take.

aomtech.ai@gmail.com
Follow us on LinkedIn