News
This page covers our work with hospitals and insurers, the programs and events we take part in, press coverage, and national policy steps on AI in healthcare.
-
We attended the 12th Turkic World Medical Congress
The 12th Turkic World Medical Congress (Türk Tıp Dünyası Kurultayı), held under the auspices of the Presidency of the Republic of Türkiye and organized by the Ministry of Health and the Health Institutes of Türkiye (TÜSEB), brought together scientists, health professionals and industry representatives from many countries in Istanbul on 29–31 October 2026. At the Congress we also had the opportunity to meet the Deputy Minister of Health.
We are developing MINA to work alongside Turkish physicians with a command of Turkish clinical language. We believe the health systems of the Turkic world, which share a common language and medical tradition, can also benefit from this approach.
TÜSEB announcement → -
The training data we prepared for MINA's new version
We shared the scale of the training data we prepared for MINA's new version. We prepared 143,000 clinical texts for pre-training, 2.7 million examples in the training set and 21,000 tool-use examples. Together, the pre-training text and the training examples contain more than 1.6 billion tokens. We kept the 13,500 questions in 98 evaluation sets separate from the training data.
Read the article → -
MINA answered 92.1% of TUS questions correctly
We measured MINA with 290 questions from Türkiye's Medical Specialty Exam (TUS) published by ÖSYM. MINA answered without tools, from its own knowledge alone, and got 267 right: 93.9% in basic medical sciences and 90.2% in clinical medical sciences.
Read the article → -
We train MINA's weights ourselves
We train MINA's new version in two stages. The model first reads clinical guidelines, drug information and the literature to build clinical knowledge into its weights, and then learns to reason over cases and state its decision with its rationale and basis. Our 17-person team works on MINA, and our board of physicians from 15 specialties approves model updates.
Read the article → -
TIS brings a patient's records together in one digital twin
A patient's past visits, lab results, imaging reports and discharge summaries sit in different systems. TIS brings these records together in the patient's digital twin, a single live profile. MINA reads this profile and presents differential diagnosis suggestions, the appropriate ICD-10 code and drug contraindications to the physician with their rationale and source. The physician always makes the diagnosis.
Read the article → -
MINA answered 98.9% of clinical knowledge questions correctly
We measured MINA with 276 questions on laboratory tests, SGK's rational test ordering rules, drug and test information, and pre-prescription checks. MINA answered 273 of them correctly and responded in 4.8 seconds on average. The physician always makes the decision.
Read the article → -
Expertise lives in the model, not the document
In general-purpose models, expertise often sits in the uploaded document; in MINA, it is in the model itself. Turkish clinical language is in the weights and expertise in 14 specialties in LoRA adapters, regulations are applied through dated rules, and the same model works in the clinical assistant, on the TIS and SIT platforms and on the digital twin. On a set of 1,451 regulation and rule questions, 98.1% were answered correctly.
Read the article → Read on LinkedIn → -
We measured MINA in four roles: auditor and hospital reached the same finding 95.9% of the time
We had the same healthcare files read from the auditor's, the hospital's, the public assessor's and the physician's point of view, and measured MINA across 510 questions. An answer counted as correct only when both the decision and the rule behind it were right. Results: auditor 94.5%, hospital 93.1%, public assessment 99.4%, physician 100%.
The auditor and hospital roles reached the same finding on 95.9% of the same files, so a hospital can see what the auditor will see, with the same rule and basis, before issuing the invoice. The results are from an internal evaluation on synthetic and anonymized scenarios.
See the results → -
Opinion AI is among the startups admitted to Terminal Istanbul
The former Atatürk Airport building has opened its doors to technology startups as Terminal Istanbul. Opinion AI is one of the startups admitted, and at the opening Ecosystem Meetup we shared the venue with technology companies from Türkiye and around the world.
We have been developing healthcare AI since 2021, and today we carry that work into the field alongside hospitals and insurers. With our Terminal Istanbul branch, we aim to take this work from Türkiye to the world.
Read on LinkedIn → -
Presenting our work on Turkish clinical data at GITEX AI Türkiye
Our stand at GITEX AI Türkiye in Istanbul drew strong interest. Among its visitors were Mehmet Fatih Kacır, Minister of Industry and Technology; Sadullah Uzun, Director General of National Technology and Artificial Intelligence; and Furkan Karayaka, Vice President of the Presidency of the Republic of Türkiye Investment Office. We walked them through the work we have developed with Turkish clinical data.
We see contributing to the healthcare pillar of Türkiye's Artificial Intelligence Action Plan as a responsibility.
Read on LinkedIn → -
Talking about AI in health insurance on NTV Tekno Hayat
During the HackZone by Allianz Türkiye Demo Day, we were guests on NTV Tekno Hayat. On the program, we introduced MINA, a healthcare AI five years in development and adapted to the clinical language of 14 specialties.
In the conversation, we explained that MINA is built not as a chatbot but as a solution that integrates into an institution's own systems, that it runs in a closed environment in line with KVKK and the GDPR, and that in time-critical processes it pursues speed and clinical decision accuracy together. We also said that our goal is to change how health insurance works.
-
Presenting MINA and our platforms at the HackZone by Allianz Türkiye 6th cohort Demo Day
At the 6th cohort Demo Day of HackZone by Allianz Türkiye, the Turkish insurance sector's first open innovation program, we took the stage to present our view of decision-making in healthcare, the MINA model and our platforms.
We joined the program as one of seven startups selected from nearly a hundred applications. Our work centers on data-driven, trustworthy decision intelligence where healthcare and insurance meet.
In its Demo Day coverage, eGirişim described Opinion AI as a startup that supports healthcare institutions and insurers in diagnosis and treatment processes through digital twins and big data analytics. Sigortacı Gazetesi's report on the close of the program's 6th cohort also listed Opinion AI among the seven selected startups.
-
Türkiye's Artificial Intelligence Action Plan (2026–2030) published in the Official Gazette
Türkiye's Artificial Intelligence Action Plan (2026–2030), coordinated by the Ministry of Industry and Technology, has been published in the Official Gazette (Resmî Gazete). Healthcare is among the plan's priority areas, and the principles of data sovereignty, human-centricity and trustworthy AI are now set at the level of national policy.
We submitted input during the plan's preparation. The risk-based regulation, EU alignment, anonymized data sharing and regulatory sandboxes we proposed are included in the plan.
Full story → Read on LinkedIn → -
The 2026–2030 Artificial Intelligence Action Plan announced at the Türkiye Artificial Intelligence Summit
At the Türkiye Artificial Intelligence Summit held at Tersane Istanbul, the 2026–2030 Artificial Intelligence Action Plan was announced by the President of Türkiye. Opinion AI attended the summit as an invited guest. Opinion AI had contributed input to the roadmap as an industry stakeholder.
The plan's "produce" and "govern" axes match the area we have worked on for five years. For us, these axes describe a national AI infrastructure for healthcare that delivers clinical decision support, complies with regulation and keeps data within the country.
Full story → Read on LinkedIn → -
Selected for the 2026 cohort of HackZone by Allianz Türkiye
We were one of seven startups selected for the 2026 cohort of HackZone by Allianz Türkiye, the open innovation program Allianz Türkiye runs in collaboration with Tenity.
We planned to run a joint proof of concept (POC) with Allianz teams during the program.
Full story → Read on LinkedIn → -
AXA Sigorta to use Opinion AI's clinical decision support system in pre-authorization
According to a report by Anadolu Agency, AXA Sigorta has entered into a strategic collaboration with Opinion AI. Through a system integration between the two companies, Opinion AI's AI-based clinical decision support system is being built into AXA Sigorta's pre-authorization processes. The aim is to improve decision quality and speed in pre-authorization review and to give policyholders faster, more transparent service.
The system analyzes medical necessity by checking diagnoses and requested procedures against current clinical guidelines, and supports each recommendation with citations and references from the scientific literature. Digital twin technology brings together a holistic view of patient data and the physician's decision model. The final decision is always made by AXA Sigorta's medical experts.
“This model, in which the final decision always rests with our AXA Sigorta health specialists, will both increase our operational efficiency and help us put our empathy assurance approach in healthcare into practice more strongly.”Yavuz Ölken, CEO of AXA Sigorta · Anadolu Agency, 26 November 2025 (translated)
We describe our role in the pre-authorization process in more detail on the Success Stories page.
-
Sharing our view on AI and human expertise in Enterprise Next
In an Enterprise Next article titled "Yapay Zeka Rakip mi, Ortak mı?" (Is AI a rival or a partner?), Opinion AI explained that AI and human expertise are not rivals but complementary forces that strengthen each other.
In healthcare, the final decision and the ethical responsibility remain with the physician; the role of AI is to offer physicians information grounded only in medicine, expertise and sources. This is consistent with our principle that people make the decision, while AI shows its rationale and source.
Read on Enterprise Next →
We describe our work in insurance, hospitals and public policy in more detail on the Success Stories page.
Success Stories