How eXalt Built a Secure and Scalable ChatGPT Alternative with Koyeb
eXalt is a French consulting firm with over 1200 consultants and offices in Paris, New York, London, Madrid, Lisbon, Brussels, and throughout France. They specialize in Finance and Tech, offering expertise in data science, cybersecurity, software development and IT infrastructure, project and product management, and more.
When eXalt consultants are on assignment, they often need fast, reliable access to a ChatGPT-like tool to help with research and problem-solving. However, many of eXalt’s clients have policies and preferences against using third-party AI services like ChatGPT.
That’s when Mehdi, eXalt’s Director of AI Practice, started looking for a secure, flexible way to build and deploy their own internal AI platform — and found Koyeb.
Challenge: Building a Secure and Scalable ChatGPT Alternative
For Mehdi and the team at eXalt, the value of tools like ChatGPT was clear: they streamline workflows, support consultants in real time, and improve productivity. But relying on OpenAI's hosted solution wasn’t an option.
"Many of our clients have strict policies around data privacy and external AI tools. We needed a solution we could fully control — one we could host ourselves," Mehdi explains. As data scientists, the team knew how to build models. But getting them into production was a recurring bottleneck.
"Each time we needed to deploy and scale, it was painful. Whether it was building out the architecture or managing the infrastructure, we would need a full-time DevOps engineer — and we didn’t have one.”
That’s when Mehdi discovered Koyeb.
Thanks to Koyeb, I can deploy services in a few clicks. In seconds, our platform is running on high-performance serverless GPUs. It’s fast, stress-free, and everything just works.
Deploying AI Solutions without a DevOps Team
eXalt used Koyeb to quickly spin up their internal platform — leveraging seamless one-click deployments for open-source models and tools like Ollama and Open WebUI, high-performance GPUs, and scale-to-zero to build a cost-efficient, production-ready ChatGPT alternative.
Mehdi says Koyeb's AI-first approach, smooth developer experience, intuitive interface, and clear documentation made it easy to deploy and manage their stack with confidence.
“Koyeb removes the DevOps overhead and lets us focus on what we’re good at.”
Tech stack powering their in-house ChatGPT alternative:
- Backend: Python, with LangChain for orchestration
- Model serving: Ollama
- Models: Mistral, DeepSeek, Gemini, and others depending on the use case
The backend services are containerized, then built and deployed directly on Koyeb.
The platform Mehdi runs on Nvidia L40S GPUs built is available to all of eXalt’s 1200 employees, with over 200 people using it heavily, especially during client-facing demos. Running these demos in parallel with the team’s internal usage made the eXalt team realized they needed to enable autoscaling to handle the traffic spikes.
By deploying on Koyeb, Mehdi and his team were able to eliminate need for a full-time DevOps team member and avoided the complexity of relying on other cloud providers — saving both time and resources.
Our team is great at engineering and building solutions. Handling the infrastructure and deployments would only slow us down. With Koyeb, everything - our services and managed database — is deployed in no time.
Koyeb’s focus on simplicity and speed helped them move fast, without compromising reliability. “Managing the service’s cloud infrastructure on our own would be painful. With Koyeb, everything just works,” says Mehdi. ”Fast deployments and zero stress — Koyeb is great.”
Evaluating Models Side-by-Side
Once eXalt had their internal platform up and running, it became more than just a productivity tool — it also became a powerful way to demonstrate value to clients.
“We use the platform to show clients what’s possible,” Mehdi explains. “It’s not just about building AI tools — it’s about helping them make informed decisions.”
The team uses the platform to run side-by-side comparisons of leading language models like ChatGPT, DeepSeek, Gemini, Claude, and Mistral. Clients can see real-time differences in output quality, latency, and cost — thanks to the platform they built.
“Our clients really enjoy comparing the performance and price across different models. It helps them choose what works for their use case.”
Accelerating Deployments From Weeks to Minutes
Deploying new services used to be a slow, manual, and painful process. Setting up cloud infrastructure on a traditional hyperscaler meant planning, building out the architecture, and handling DevOps tasks the team simply didn’t have the time or bandwidth for.
The deployment process could take weeks, if not longer.
With Koyeb, that entire process changed. “We can deploy in minutes — without managing the setup or any of the infrastructure,” explains Mehdi.
“In one-click, I can deploy the open source models and services I need, and they’re live in minutes,” says Mehdi. “Koyeb lets me test different models instantly and deploy services way faster than we ever could before.”
Optimizing Infrastructure and Costs with Scale to Zero
“Scale to zero is amazing,” says Mehdi. It’s one of the standout features that the eXalt team loves about Koyeb, providing both cost savings and high performance.
“When the platform isn’t in use, we don’t pay for idle resources. It’s a huge cost-saving benefit,” explains Mehdi. “As soon as we need the service, it spins back up instantly. It’s perfect.”
With serverless GPUs that scale to zero, eXalt only pays for the resources they actually use — all without avoiding the complexity of infrastructure management.
Focusing on Developing ML and AI Solutions
Mehdi leads a 16-person team focused on building and testing machine learning and generative AI solutions. Their work spans use cases, models, and custom tools — and with limited bandwidth, every hour spent on infrastructure is an hour not spent on innovation.
“We’re a team of ML engineers and data scientists — not DevOps. Koyeb lets us stay focused on what we do best.” By handling deployments, scaling, and infrastructure behind the scenes, Koyeb frees the team to concentrate on developing other projects for their team and clients.
“We leave the infrastructure to Koyeb. It just works — and that’s exactly what we need.”
Koyeb Features Highlighted:
- Serverless GPUs: Run GPU workloads on demand without managing hardware
- Scale to Zero: Automatically scale down idle services to reduce costs, then spin them back up instantly when needed
- Autoscaling: Automatically scale resources based on usage and only pay for what you need without worrying about idle time or infrastructure management
- One-Click Deployments for Ollama, Mistral 7B, DeepSeek, and more
- Managed Database: Launch managed PostgreSQL databases alongside your services without handling provisioning or maintenance
- Docker deployments: Deploy any containerized app effortlessly using your existing Dockerfiles
- Metrics: Get real-time insights into your services with built-in dashboards for usage and performance
🚀 Ready to Build the Future of AI?
eXalt is at the forefront of generative AI and machine learning, helping companies build and adopt AI solutions. If you’re excited about shaping the future of technology, eXalt is hiring!
Need cutting-edge expertise and solutions? Check out their services.