ConfidentialMind (@ConfiMind) header
ConfidentialMind (@ConfiMind) profile picture

We help organizations to own and deploy AI in secure and sovereign environments

Espoo, Finland
Joined April 2023
Photo by ConfidentialMind (@ConfiMind)Photo by ConfidentialMind (@ConfiMind)Photo by ConfidentialMind (@ConfiMind)Photo by ConfidentialMind (@ConfiMind)Photo by ConfidentialMind (@ConfiMind)Photo by ConfidentialMind (@ConfiMind)Photo by ConfidentialMind (@ConfiMind)Photo by ConfidentialMind (@ConfiMind)
We have partnered with Vivicta (linkedin.com/company/vivicta…) to build shared sovereign AI capabilities that can help lower the cost and complexity of adoption, while giving organizations the control, security and governance they need. For customers, the goal is simple: make advanced AI, from LLMs and RAG to agentic solutions, easier to deploy and operate in secure and trusted environments. More control shouldn’t mean less innovation. 👉 Read how Vivicta and ConfidentialMind are making sovereign AI more practical across the Nordics. bit.ly/48pZVhO #AI #LLM #SOVEREIGN #SecureAI
1
2
4
48
Your next AI service could run on the GPUs you already own. At @Supercomputing 26, we’re announcing #Slurm support in ConfidentialMind. Organizations can now deploy models onto connected Slurm clusters and use them through the same ConfidentialMind AI platform that supports dedicated GPU deployments - with retrieval, agents, chat, governance, tenant controls, and usage metering. For universities, national labs, and research-driven organizations, this opens up practical opportunities:  • Share model deployments across users and projects, reducing duplication.  • Evaluate new models and add inference capacity without a separate hardware purchase.  • Access a larger GPU pool for higher availability and more flexible scalability than a single dedicated node.  • Start on Slurm, combine it with dedicated GPUs, or keep model serving within the HPC cluster. It also supports a shared operating model: Research IT manages the Slurm infrastructure, while central IT manages the platform and dedicated capacity. Both can serve their communities without each building and maintaining a separate AI service. With tight budgets and long procurement lead times, making better use of existing infrastructure matters. Come meet us at booth 6815 - SC26 in Chicago, November 17–19. Link to the full article: confidentialmind.com/blog/sl… #SC26 #HPC #Slurm #AIInfrastructure
1
2
2
63
This week we’re heading to #SC26 in Chicago - and we’re proud to be featured in Intel Innovation Pass Program at @Supercomputing 2026! At #SC26, we’re demoing the ConfidentialMind Private AI Platform, running live on Intel® Xeon 6741P processors and Intel® Arc Graphics, bringing secure, enterprise-grade agentic AI to infrastructure you control. And we’re engineering for the next scale: the same ConfidentialMind platform running open Intel® Xe software stack carry forward to the upcoming Intel® GPU D-series. Build once, scale into the next generation. Come see it live at SC26 - Booth 6815. Let’s talk about what private agentic AI can do for your organization. #SC26 #AgenticAI #IntelXeon #PrivateAI
2
2
92
We have partnered with @tenstorrent to deliver Sovereign AI on Open Hardware. We are excited to partner with Tenstorrent, bringing our self-hosted AI platform to Tenstorrent Galaxy Blackhole systems. Customers get LLMs, RAG, and agents running on open silicon, in their own data center, without giving up control of their data. We’re selling the combined stack together across Europe, the Middle East, Asia-Pacific, and the Americas. Read more here: confidentialmind.com/blog/te… See us at #SC26, booth 6815!
We’ve partnered with @ConfiMind to bring its self-hosted AI platform to Tenstorrent Galaxy. Run open source LLMs, RAG, and agents in your own data center, without having to build a platform first. Learn more → bit.ly/4zlvDry
1
2
4
129
📬 Postcards from @StockholmMLOps #40! We spent wednesday evening inside Sweden’s first nuclear reactor, 22 meters underground at @KTHuniversity . What an event! Our Senior AI Research Engineer @specfinaj spoke about SampoTron, our query-side router for cross-lingual retrieval for Nordic languages. We were in good company on stage with CHI Software , @hopsworks , and Sweden AI Factory. The energy in the room matched the venue. We were impressed with the deep expertise in the room across engineering, AI, ML, and infrastructure to mention a few. Conversations wandered from retrieval to self-hosting, infrastructure and agentic coding. Throw it all together and there’s a good read on where Nordic AI actually stands right now. Thanks to Stockholm MLOps and their welcoming community, CHI Software, Hopsworks, and Sweden AI factory for a great event and making this possible!
2
5
116
We have had a wonderful and productive time showcasing our private AI solution @NVIDIAGTC in Berlin this week. If you are looking at owning your AI, your agents without data leaving your premises, then do get in touch with us!
2
4
63
ConfidentialMind retweeted
Confidential Mind from Finland @ConfiMind does the same as Azure Foundry and AWS Bedrock, offering simple APIs to use LLMs, but for local models. Many university customers already.
1
1
6
481
Last week our CEO @MRasanen had the opportunity to be part of a great evening in Stockholm with @JensenHuang and the rest of the #Nordic AI community. It was an honor and a matter of great pride that we were the only Finnish early stage company featured. If you're looking to own your own agentic AI without the data leaving your premises, then do get in touch with us at @ConfiMind . Now back to shipping awesome #private AI to our wonderful customers.
3
76
We still can’t get over SampoTron, our SOTA cross-lingual encoder. Built through post training of NVIDIA Nemotron-3-Embed-1B, SampoTron improves average cross-language retrieval ranking quality by 20.9% over the original model without rebuilding the document index. We’re helping enterprises own their AI without making language a barrier to finding knowledge. Your workforce and agents should be able to find relevant information even when your documents are written in another language.
2
5
133
Introducing SampoTron: Our SOTA Cross-lingual Embedder Cross-language retrieval is a real limitation in multilingual RAG: a Finnish query needs to surface an English passage, but most encoders don’t bridge that gap well. Fixing it usually means re-embedding your entire document index. We built SampoTron to get around that. It’s a query-only LoRA adapter paired with a language router, trained on top of @NVIDIAAI Nemotron-3-Embed-1B-BF16. Result: 20%+ relative improvement in cross-lingual retrieval accuracy (nDCG@10) across English, Finnish, and Swedish. All six language-pair directions improved, with native-language retrieval unaffected. The NVIDIA stack runs end to end: training and translation data preparation utilized an on-prem NVIDIA B300 GPU cluster, followed by fine-tuning on an NVIDIA RTX PRO 6000 Blackwell GPU. Subsequently, the complete production pipeline utilized SampoTron, based on the NVIDIA Nemotron-3-Embed, for retrieval, NVIDIA Nemotron 3.5 Lightning 30B-A3B NVFP4 for generation running on a single NVIDIA GB10 Grace Blackwell Superchip system.
2
3
1
7
170
ConfidentialMind retweeted
Thank you to @MRasanen, CEO of @ConfiMind for sitting down with us and sharing his thoughts on: - Whether open source models are undervalued - Changes in AI infrastructure & inference - Power consumption & efficiency in Europe - The future of open source AI
2
5
31,875
We recently made our documentation public. If you are interested in an opinionated AI inference platform that allows you to deploy complete secure AI systems (RAG, Graph RAG, etc.) to any environment (VPC, on-prem, private cloud, etc.) - go check it out: docs.confidentialmind.com/
2
4
190
Introducing Secure Copilot for R&D 1. Automatic context retrieval for code understanding. 2. Eliminate copy-pasting; Easily link to tickets/design files. 3. Boost productivity with custom code generation tailored to your needs. #GenerativeAI #Copilot #Security
1
2
241
Check out our latest blog post on the ConfidentialMind secure chat interface. Run LLMs in your own environment without the fear of data leaks. confidentialmind.com/post/pl…
1
201
ConfidentialMind retweeted
Recently, we decided to launch a new startup company in the area of helping companies utilize open source AI models. Go check it out here🔒🧠 @ConfiMind confidentialmind.com/post/co…
1
3
287