PatientSeek launches as the first open-source MED-LEGAL reasoning model powered by Deepseek

In a world where law and healthcare increasingly intersect, a new AI model is stepping in to bridge the gap. Enter PatientSeek — the first open-source MED-LEGAL reasoning model — designed to bring sophisticated medical and legal insights to professionals who need airtight correlations between patient records and external events.
Developed with meticulous attention to data structure and reasoning, PatientSeek is a specialized tool built to handle the complex demands of MED-LEGAL workflows with precision and security.
How does it work?
At its core, PatientSeek is a fine-tuned DeepSeek R1 model, trained on one of the largest accessible datasets of patient records. The team behind it painstakingly preprocessed thousands of patient histories, aligning them with the nuanced needs of MED-LEGAL analysis. Unlike traditional AI models that simply spit out results, PatientSeek is designed to reason — drawing meaningful connections between medical data and legal causation questions.
The model was developed using CometML for dataset storage and experiment tracking, Unsloth and HuggingFace TRL for fine-tuning, and AWS Sagemaker — with support from NVIDIA Inception credits — to bring this vision to life.
Security and privacy are at the forefront of this development. Organizations handling sensitive patient data can run PatientSeek offline, ensuring that critical information stays local. Performance-wise, it rivals top-tier models while keeping costs significantly lower. In fact, DeepSeek R1 as an API is 27 times cheaper than GPT-4o1, making PatientSeek an affordable yet powerful alternative.
Why does it matter?
The MED-LEGAL field requires more than just data retrieval — it demands deep reasoning. Healthcare providers and legal professionals often have to establish airtight correlations between medical conditions and external events, a challenge that traditional workflows struggle to meet.
"We realized that MED-LEGAL workflows needed a reasoning layer that could withstand legal scrutiny, while also providing the medical clarity professionals rely on," says the development team.
With expertise spanning both law and healthcare, they saw an opportunity to build a model that could handle causation questions at the level required for legal standards.
Beyond causation analysis, PatientSeek is a game-changer for automating repetitive tasks. Need a quick patient history or a reasoned answer about a patient's medication use? The model can provide context that enhances real-time patient interactions. It doesn't replace human expertise — it enhances it.
The context
The release of DeepSeek R1 and the broader industry acceptance of automated reasoning set the stage for PatientSeek's arrival. AI models no longer need to be generalized; instead, they can be refined to fit specific, high-stakes use cases like MED-LEGAL workflows.
Fine-tuning an AI model isn't just about feeding it data. Structured, high-quality data processing is the secret sauce behind performance. PatientSeek's developers meticulously formatted its training data to reflect real-world business objectives, ensuring that the model understands the complexities of medical reasoning.
The results speak for themselves. In basic clinical tasks like condition detection and vital signs analysis, PatientSeek hits around 90% accuracy — beating most open-source models and matching GPT-4o1's performance. The real magic happens when the complexity ramps up. Whether it's generating patient summaries or treatment plans, PatientSeek retains its 90% accuracy, whereas other models begin to falter.
"While general-purpose models can handle the basics, specialized models like PatientSeek truly shine in complex reasoning tasks," say its developers. This distinction makes it invaluable for medical professionals looking to streamline documentation and automate high-level analysis without sacrificing accuracy.
And it's affordable. Running on AWS costs less than a cent per basic query, while more complex reasoning tasks cost under $0.05. Plus, it can even be deployed on M2 Macs using Ollama, making it functionally free in certain setups.
With PatientSeek, MED-LEGAL professionals now have an AI ally that understands their unique challenges, adapts to their needs, and brings cutting-edge reasoning into their daily workflows — all while keeping costs and security under control.
💡Did you know?
You can take your DHArab experience to the next level with our Premium Membership.👉 Click here to learn more
🛠️Featured tool
monday
Built for a new way of working, monday is a stunning project management tool that enables the whole team to collaborate, share updates, see the performance, comments, and more.
👉 Click here to learn more
