HangiKredi took its Financial Assistant live without data leaving Turkey
HangiKredi, which compares financial products for millions of users, runs its agentic AI infrastructure and its Financial Assistant product on FixCloud’s LLMaaS and GPUaaS services. Orkun Sevinç, who leads the Data Management and AI function at HangiKredi, told us how it went.
- Company
- HangiKredi
- Financial product comparison platform
- Use case
- Financial Assistant
- Agentic AI infrastructure
- FixCloud services
- LLMaaS + GPUaaS
- One model API and GPU infrastructure
- Data location
- Turkey
- KVKK and the Banking Law
Understanding with data, personalizing with AI
HangiKredi serves millions of users, and each of them has different needs and different financial circumstances. For Orkun Sevinç, data is what makes these differences understandable, and AI is what turns that understanding into an experience that fits the person. Ranking products, personalization, making financial content easy to understand and improving operations are all parts of the same approach. The goal is an experience in which users can weigh their options better and make more informed financial decisions.
The challenge: regulation and sustainable scale
The decision to scale the AI projects brought two issues to the front. The first is regulation: KVKK (the Turkish data protection law) and the Banking Law prohibit data from leaving the country, entirely or in part. The second is running the applications sustainably: as usage grows, response time, concurrent requests, GPU capacity and cost all enter the same equation, while the world of models keeps changing fast.
“For us the critical question was how to reach the computing power we needed while keeping cost predictable and protecting the team’s development speed.”
Keeping data in Turkey means control over the data and auditability for HangiKredi. Sevinç sees KVKK compliance as a holistic responsibility that covers, along with the location of the data, access rights, record management, data minimization and contractual responsibilities.
Why FixCloud?
Three things stood out in HangiKredi’s evaluation: control over the data, access to the AI capacity it needed, and operating that setup sustainably. The decision itself was made on measurements.
“Before choosing FixCloud we ran a great many tests and POCs with a large number of vendors, and we made our decision on the performance metrics we got from them.”
The other points Sevinç highlights: fast action on questions and problems, transparent and sincere communication, and high availability.
The solution: LLMaaS and GPUaaS together
LLMaaS
Setting up large language models, serving them and keeping them running takes separate expertise and effort. FixCloud LLMaaS offers this layer as a service by giving access to different models through one common API. The team puts its focus on the use case, on working with the right data and on output quality, and it can evaluate different models against the quality, speed and cost balance of each scenario.
GPUaaS
A fixed and predictable cost, easy integration into HangiKredi’s own architecture and not having to deal with other services let the team concentrate on its core work and increase its development speed.
“Keeping the data in Turkey was more than a preference for us, because it is a responsibility in our sector and because it gives us more freedom to act. In fact, it gave us the comfort of on-prem without having to bear the DevOps cost.”
Integration: from internal use to live in a few months
The integration started with the agentic AI infrastructure and the Financial Assistant. After development and the building of guardrails, the product went live first inside the company and then outside it; the process was completed within a few months. For Sevinç, the most important side of the integration for the team was being able to work with the comfort of on-prem.
The result: high speed and data privacy
Today HangiKredi uses FixCloud infrastructure for its Financial Assistant product.
“With the Financial Assistant we used to face latency and uncertainty about how data was stored and processed; after this collaboration we reached a high level of speed and data privacy.”
Because the infrastructure work of serving models is handled as a managed service, the team can spend its time on data quality, model evaluations, integrating the application into business processes and the user experience. What stays with Sevinç from the collaboration: fast responses, sharing the importance and the responsibility of the work together, high availability, and all of it at a reasonable cost.
What comes next?
HangiKredi’s priority is to deepen personalization and to build an experience that understands the user’s financial needs more completely; the Financial Assistant is an important part of that vision. Inside the company, the aim is to support data-driven decision making and repetitive operations with AI. Sevinç puts accuracy, explainability, user control and measurable benefit at the center of this development.
Advice for companies planning an AI investment
Orkun Sevinç’s recommendations for companies that will decide on infrastructure, especially in the financial sector:
- First make the problem to be solved and the measure of success clear.
- Choose the infrastructure by data sensitivity, expected usage volume, response time and the quality you need.
- In the cost calculation, add the team’s operating burden to the hardware or service fee.
- Start small with real workloads and measure the results; the right scaling decision becomes easier.
- Evaluate the flexibility to change models, data portability and the terms for leaving the service from the start.
- Make security and governance part of the work from the first day of the design.
“At the heart of our collaboration with FixCloud is turning our team’s AI capability, on a reliable infrastructure, into solutions that bring tangible benefit to our users and our business.”