DevisionX

Home
Egypt
Cairo
DevisionX

DevisionX your end-to-end AI solutions partner.

28/10/2026

Your agent works in the demo. Production is another story.

A pod restarts and it runs the action twice. A tool call hangs, and nothing notices. It loops quietly and burns the GPU budget. A prompt injection reaches a tool with write access. And nobody can explain why it did what it did.
AgentOps for Private LLM Systems is five weeks on running AI agents as production systems instead of demos: stateful, resumable workflows on a private LLM, sandboxed tools, least-privilege permissions, human approval gates and full tracing.
You finish with a governed agent running real tools on infrastructure you control, plus the dashboards and runbook to operate it.

Starts Friday 20 November · 5 weeks · 4 live sessions https://zomra.io/courses/120ae729-e8ba-46b4-9abd-88b4044e8c22

19/10/2026

In 1994, Bill Gates said: "Banking is necessary, banks are not."

30 years of fintech later, we think the opposite is happening 👇

🏦 The bank stays: its licence, balance sheet and trust.
📄 The paperwork goes: one shared AI layer takes on the reading, checking and decisions that keep files waiting for days.

Our new white paper, Banks Without Banking, explains how.

📥 Read it free: https://banking.devisionx.com/banks-without-banking-white-paper

08/10/2026

Tonight, 8:00 PM Cairo.

90 minutes on the layer between your model weights and your application: how inference engines work, how vLLM, TensorRT-LLM, SGLang and llama.cpp compare, and how to choose and configure one from a single GPU to a production cluster.

Led by our Co-Founder & CTO Mohamed Rashad, hosted by MLOps MENA Community.

Free. Live. Online. https://zomra.io/free-sessions/e869921f-4df7-4168-adc0-7495e9b059d2

04/10/2026

Between the model weights and your application sits a layer most teams never look at until it's the reason their latency doubled.

That layer is the inference engine. Choosing, configuring and scaling one is now one of the most consequential decisions in any LLM deployment, and one of the least understood.

On Thursday, our Co-Founder & CTO Mohamed Rashad is giving a free 90-minute live session on exactly that, hosted by MLOps MENA Community

Introduction to LLM Inference Engines

→ How engines actually work: KV-cache management, continuous batching, speculative decoding, tensor parallelism, quantization → vLLM, TensorRT-LLM, SGLang and llama.cpp compared — what each is good at, and where each one hurts → How to choose and configure an engine, from a single GPU to a production cluster

Thursday 8 October · 8:00 PM Cairo time, live online

Register on Zomra: https://zomra.io/free-sessions/e869921f-4df7-4168-adc0-7495e9b059d2

30/09/2026

Last day for early-bird pricing on Advanced On-Premises LLM Infrastructure.

5 weeks. 4 live sessions. Starts 16 October. You finish with a private LLM serving platform running on your own infrastructure, built by you.

If your AI has to stay inside the building, this is the engineering that makes that survivable.

https://zomra.io/courses/advanced-on-premises-llm-infrastructure-bm8a7x

24/09/2026

DevisionX is exhibiting at the IEEE Global Career Fair 🚀
We're building AI that sees, thinks, and acts, and we're looking for people who want to build it with us.

Come by our virtual booth to:
🔹 Meet the team behind Tuba.AI, our no-code AI Vision workflow builder, and our Multimodal RAG Assistant
🔹 Learn how we automate inspection, monitoring, and decision-making for regulated industries like BFSI
🔹 Talk about open roles and internships in Computer Vision, Document Intelligence, LLMs/RAG, and MLOps

If you'd rather ship real AI to production than keep it stuck in a notebook, we'd like to meet you.

📅 Today 24 September 2026
📍 Virtual: IEEE Global Career Fair → search "DevisionX" in the Exhibit Hall
🔗 https://lnkd.in/d9C2dUeU

See you there 👋

23/09/2026

AI Made in Africa and DevisionX was in the room.

Last week in Berlin, our Founder & CEO Mahmoud AbdelAziz represented DevisionX at "Invest in African AI: From Innovation to Investment" — a side event on GIZ African Union AI Made in Africa track at the BIG BANG KI FESTIVAL.

The theme of the week was simple, and overdue: Africa is not only a market for AI built elsewhere. It's a place where AI is being built.
GIZ's own mapping puts numbers behind it — 300+ AI businesses across 38 African countries, already deployed in agriculture, health, finance and public services.

DevisionX is one of them, we build AI that sees, thinks and acts — computer vision, document intelligence and multimodal RAG — deployed inside regulated industries where accuracy, privacy and on-prem control aren't features, they're requirements.

What we took from Berlin: European enterprises and integrators aren't looking for charity projects. They're looking for partners with real deployments, hard constraints already solved, and teams that ship. That's a conversation African AI companies are ready to have as equals.

Thank you to GIZ and the AI Made in Africa team for building the bridge and to everyone who stopped by to talk vision, RAG and what "trustworthy AI" actually costs to deliver.

More on the initiative and the new report from across the continent:
🔗 https://www.giz.de/en/regions/africa/african-union/news/ai-made-africa-voices-ai-businesses-across-africa

22/09/2026

Four weeks. Four problems that decide whether a private LLM deployment survives contact with production.

01 · Private LLM Infrastructure Architecture Workload sizing, compute capacity planning, the physical layer, plane separation, deployment topologies, high availability. Get this wrong and no amount of tuning later saves you.

02 · Advanced Multi-GPU Inference vLLM runtime, tensor/pipeline/data parallelism, continuous batching, KV-cache management, prefix caching, quantization. This is where most teams find out their throughput numbers were fiction.

03 · Production LLM Serving Platforms OpenAI-compatible gateway, request routing, multi-model serving, workload isolation, authentication, quotas, rate limits. The difference between a demo and a platform other teams can build on.

04 · Performance and Reliability Engineering Benchmarking, load testing, GPU telemetry, SLO definition, alerting, recovery procedures.

Then you build the whole thing: a complete private LLM serving platform, yours to keep.

Starts 16 October. 5 weeks, 4 live sessions. Early-bird closes 30 September. https://zomra.io/courses/advanced-on-premises-llm-infrastructure-bm8a7x

20/09/2026

Most engineers learn to build an LLM application. Very few learn to run one.
That gap is the whole job. Inside a bank's perimeter, on your own GPUs, the questions stop being about prompts and start being about multi-GPU inference, request routing, KV-cache and quantization, GPU telemetry, SLOs and what happens at 3am when a node dies.

It's the work we do at DevisionX for clients in BFSI and other regulated sectors. We've now put it into two live courses, in partnership with MLOps MENA Community, running on Zomra.

1- Advanced On-Premises LLM Infrastructure · starts 16 October
Private LLM infrastructure architecture · advanced multi-GPU inference with vLLM · production serving platforms and request routing · performance and reliability engineering.

→ You finish by building a complete private LLM serving platform.
https://zomra.io/courses/advanced-on-premises-llm-infrastructure-bm8a7x

2- AgentOps for Private LLM Systems · starts 20 November
Production agent architecture with stateful orchestration and MCP tool contracts · secure tool ex*****on and governance · agent evaluation and observability · runtime reliability.

→ You finish by building a governed on-premises AI agent.
https://zomra.io/courses/120ae729-e8ba-46b4-9abd-88b4044e8c22

Each course runs 5 weeks with 4 live sessions, and each stands on its own, take one or both.

Taught by Mohamed Rashad, our Co-Founder & CTO

Early-bird pricing is open now and closes soon. If on-prem is where your AI has to live, enroll while it's still on.

06/09/2026

The inaugural Cyber4Africa cohort represents far more than 49 promising startups.

Together, our fellow innovators and we represent a new generation of African AI innovators building solutions in healthcare, fintech, govtech, energy, and AI infrastructure while tackling increasingly complex cybersecurity challenges.

As startups, we are not starting from zero. We already understand the importance of cybersecurity. What we need is practical support to strengthen and formalize the security foundations we have already built. We want cybersecurity to become a catalyst for growth, enabling them to scale into new markets and unlock larger commercial opportunities.

Over the coming months, we will work to:
- Strengthen cybersecurity governance to protect sensitive data
- Build security by design into AI systems from the outset
- Develop the internal capacity needed for long-term resilience
- Meet international standards and certification requirements

A huge thank you to AI Hub for Sustainable Development , UNDP, UNDP Digital, AI and Innovation Hub (Accelerator Labs), Ministry of Enterprises and Made in Italy , Cisco, Cisco Networking Academy , CYBER 4.0 - Cybersecurity Competence Center, and everyone else who has supported this vision.

We're excited to begin our journey as we help shape a more secure and trusted AI ecosystem across Africa.

Learn more: https://aihubfordevelopment.org/cyber4africa-inaugural-cohort

Address

124 Othman Ibn Affan
Cairo
11361

Website

https://devisionx.com/

Alerts

Be the first to know and let us send you an email when DevisionX posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.