Our COO Oleg Greshnev pitched a simple but critical idea: GenAI can reason, but it cannot calculate. Finmars is the verified financial data layer that fixes that, giving AI agents numbers they can trust and boundaries they cannot cross.
Our collaboration with SnT, Interdisciplinary Centre for Security, Reliability and Trust / University of Luxembourg brings this into asset allocation, automating company analysis with DCF and credit models built into the workflow.
It was an inspiring morning of sharp presentations, meaningful conversations, and a real sense of where the industry is heading with Generative AI.
Weโre proud to have represented Finmars alongside Polina Kravchenko and Alexander Yakubov, and grateful for every conversation had at the mini-fair and networking lunch.
The momentum in the Luxembourg AI ecosystem is real. We’re excited to be part of it. ]]>
2025 was about turning Finmars into something people can actually use, fork, and build on.We werenโt sitting around this year and worked hard for moving forward towards Open Finance world.
๐ช๐ต๐ฎ๐ ๐๐ฒ ๐ฑ๐ฒ๐น๐ถ๐๐ฒ๐ฟ๐ฒ๐ฑ:
โข ๐ช๐ฒ ๐ฏ๐ฒ๐ฐ๐ฎ๐บ๐ฒ ๐๐ต๐ฒ ๐ณ๐ถ๐ฟ๐๐ ๐ผ๐ฝ๐ฒ๐ป-๐๐ผ๐๐ฟ๐ฐ๐ฒ ๐ณ๐ผ๐๐ป๐ฑ๐ฎ๐๐ถ๐ผ๐ป ๐ณ๐ผ๐ฟ ๐ผ๐ฝ๐ฒ๐ป ๐ณ๐ถ๐ป๐ฎ๐ป๐ฐ๐ฒ:https://lnkd.in/dfNi33uU
Deploy a full PMS and populate it with data in a day, not months. Run it in the cloud, on-prem, or even on your laptop. Full data control. No vendor lock-in. Unlimited flexibility
โข ๐๐ฃ๐ ๐ฑ๐ผ๐ฐ๐๐บ๐ฒ๐ป๐๐ฎ๐๐ถ๐ผ๐ป
Clearly structured APIs to integrate Finmars into existing banking, PMS, and data environments
โข ๐จ๐๐ฒ๐ฟ-๐ณ๐ฟ๐ถ๐ฒ๐ป๐ฑ๐น๐ ๐๐ป๐๐๐ฎ๐น๐น๐ฎ๐๐ถ๐ผ๐ป ๐ด๐๐ถ๐ฑ๐ฒ๐
Practical documentation that lets teams install Finmarsย for free, validate use cases, and evaluate the platform before any commercial commitment
โข ๐ ๐ฎ๐ฟ๐ธ๐ฒ๐๐ฝ๐น๐ฎ๐ฐ๐ฒ ๐น๐ฎ๐๐ป๐ฐ๐ต
A modular extension layer for Finmars, enabling add-ons, integrations like Exante, and partner solutions. Tool for developers to monetize their solutions
โข ๐ฃ๐ ๐ฆ ๐ฎ๐ฑ๐ฑ-๐ผ๐ป ๐๐ฟ๐ฒ๐ฒ & ๐ฃ๐ฟ๐ฒ๐บ๐ถ๐๐บ ๐ฉ๐ฒ๐ฟ๐๐ถ๐ผ๐ป๐
Portfolio Management Solution available via the Marketplace and built for real production workflows. Available on laptop and mobile
โข ๐๐ฎ๐๐-๐๐ผ-๐๐๐ฒ ๐๐ฒ๐บ๐ฝ๐น๐ฎ๐๐ฒ๐ ๐ฑ๐ฒ๐๐ถ๐ด๐ป๐ฒ๐ฑ ๐ณ๐ผ๐ฟ ๐๐ถ๐ป๐บ๐ฎ๐ฟ๐ ๐ฃ๐ ๐ฆ
Upload, validate, and start working with portfolios, transactions, and reports in minutes with no complex setup required
โข ๐ฅ&๐ ๐๐ถ๐๐ต ๐๐ ๐ณ๐ผ๐ฐ๐๐
Ongoing R&D projects withย SnT / University of Luxembourg, including work around AI-assisted analytics, configuration, and future automation inside the platform
We believe that it is exactly what open finance should be in the era of AI: open, yet secure and extendible via customization tools and marketplace.
Thanks to everyone building, testing, and challenging us along the way!
If youโre working on fintech, data, or open-source infrastructure โ letโs talk.
Already looking ahead for 2026?
๐๐ง๐ ๐๐ฅ๐๐๐ซ ๐ญ๐ก๐๐ฆ๐ ๐๐ซ๐จ๐ฆ ๐ญ๐ก๐ข๐ฌ ๐๐ฏ๐๐ง๐ญ ๐ข๐ฌย ๐ญ๐ก๐ ๐ข๐ง๐๐ฎ๐ฌ๐ญ๐ซ๐ฒ ๐ข๐ฌ ๐ฆ๐จ๐ฏ๐ข๐ง๐ ๐๐๐ฌ๐ญ ๐ญ๐จ๐ฐ๐๐ซ๐ ๐๐-๐ง๐๐ญ๐ข๐ฏ๐ ๐ข๐ง๐๐ซ๐๐ฌ๐ญ๐ซ๐ฎ๐๐ญ๐ฎ๐ซ๐: production pipelines over PoCs, AI as a foundation layer, full-stack redesigns, unpredictable outcomes that still demand investment, and sovereignty becoming a hard engineering requirement.
๐๐ญ ๐ ๐ข๐ง๐ฆ๐๐ซ๐ฌ, ๐๐ ๐ข๐ฌ ๐ง๐จ๐ญ ๐ ๐๐ฎ๐ณ๐ณ๐ฐ๐จ๐ซ๐: itโs a core part of our joint research with SnT, Interdisciplinary Centre for Security, Reliability and Trust, where we are developing scientific components that make financial infrastructure more intelligent and more accessible.
Together with SnT, we are working on:
โขย ๐๐ ๐๐ฌ๐ฌ๐ข๐ฌ๐ญ๐๐ง๐ญย that can understand user intent, answer portfolio questions, generate configuration help, and provide proactive alerts based on portfolio data
โขย ๐๐๐-๐๐ซ๐ข๐ฏ๐๐ง ๐ฆ๐๐ซ๐ค๐๐ญ ๐ง๐๐ฐ๐ฌ ๐๐ง๐๐ฅ๐ฒ๐ฌ๐ข๐ฌย to identify market movers, extract key theses, and understand how narratives impact future market states
โขย ๐๐ฎ๐ญ๐จ๐ฆ๐๐ญ๐ข๐ ๐ฉ๐จ๐ซ๐ญ๐๐จ๐ฅ๐ข๐จ ๐๐ฅ๐ฅ๐จ๐๐๐ญ๐ข๐จ๐ง ๐๐จ-๐ฉ๐ข๐ฅ๐จ๐ญ, able to recompute allocations based on user constraints, market states, expert views and NLP-identified investment ideas
These research activities are part of a collaboration that advances AI in asset allocation, financial analytics, and user assistance with results becoming open-source and integrated into the Finmars platform.
]]>
Finmarsย showcased how open-source infrastructure is turning AI concepts into measurable results โ faster reporting, adaptive risk, and leaner operations across Luxembourgโs funds ecosystem.
A huge thank-you to the organizers and to all the brilliant speakers and peers who made the session so actionable and inspiring!
ย If you believe finance should beย open, smart, and sovereign, explore our platform on GitHub โ build, fork, or run your own modules today:
https://lnkd.in/dfNi33uU
]]>We believe that for this generation to unlock their full potential and bring bold ideas to life, they needย open technologies and communities. Thatโs why Finmars is open source, so anyone can build their own modules, contribute new ideas, and shape the future of finance together.
A big thank you to Elisa Jensen Labiano, Antony Martini and the LSB ร The LHoFT – Luxembourg House of Financial Technology team for making this happen!
ย Curious how Finmars is building the future of open finance? Explore and try our platform on GitHub: https://lnkd.in/dfNi33uU
]]>The Challenge: Low Innovation Adoption in Investment Management
Despite intense buzz around artificial intelligence and fintech innovation, the actual adoption of next-generation models remains surprisingly low in the investment management spaceโespecially in traditional asset management. Many institutions still rely heavily on Excel, legacy modelling tools, or internally built black-box systems that donโt scale, lack modularity, or resist extension.
As a result, trillions of dollars are allocated based on static or semi-static models that struggle with nonlinear markets, data complexity, and evolving risk regimes. Leading consultancies like McKinsey estimate that AI, generative AI, and agentic AI could unlock value equal to 25โ40% of an asset managerโs cost base. (McKinsey & Company) Meanwhile, BCGโs surveys of firms managing over USD 15 trillion find that two-thirds either plan or are already scaling generative AI use cases. (BCG)
Yet many firms hesitateโnot because the value is unclear, but because the path is. The obstacles are real:
- Lack of shared, open standardsย that allow interoperability, extension, and community innovation
- High integration and maintenance costsย for bespoke AI solutions
- Slow governance, auditability, explainability, and regulatory constraints
- Fragmented tooling and data silos
These barriers keep promising models from moving out of the lab and into production.
What Weโre Doing Together: Research + Platform + Delivery
In our joint collaboration, we aim to span the full โinnovation value chainโ:
| Stakeholder | Role & Contribution |
| SnT | Advances foundational AI research; leads prototyping and proof-of-concept work |
| FINMARS | Provides the technical backbone: open architecture, APIs, SDKs, deployment tooling, and conduit to financial domain models and production-ready templates |
| End Users / Early Adopters | Participate in validating prototypes, pilot deployments, feedback loops, extensions, and ecosystem contributions |
Some highlights to date:
- Several successful PoCs integrating domain-driven models (e.g. factor enhancements, sentiment-based overlays)
- Architecture designs that allow such models to be packaged and deployed modularly
- Early planning of a โSolutions Marketplaceโ through which developers, asset managers, and partners can share and distribute model plugins
Going forward, the roadmap includes:
- Pilot scaling: Working with pilot customers to deploy modules under real-world conditions
- Ecosystem seeding: Incentivizing third-party developers and researchers to publish modules, plugins, and enhancements
- Governance & compliance tooling: Embedding explainability, audit trails, version control, and risk metadata
- Community & education: Building documentation, tutorials, workshops, and open reference models
Why Finmarsโs Approach Matters
“By offering anย open-source, API-first, AI-friendly, modular platform, Finmars lowers the barrier to entry for both financial firms and technical teams. Instead of reinventing the wheel, development teams can start with baseline building blocks, optimize or extend them, and publish to a shared marketplace. This model magnifies innovation and shortens time-to-market.
Our vision is that over time, this shared ecosystem becomes theย default substrateย for financial innovation โ much like how certain open standards powered the mobile, web, or cloud ecosystems.”
Oleg GRESHNEV, COO at Finmars
โWorking with SnT has sharpened our view of how academic innovation must meet industrial rigor. Together, weโre laying the foundation for AI models that are reliable, extensible, and usable in production,โ saidย Alexander Yakubov, Research Lead at Finmars.
Call to Action
If you are:
- aย financial institution or asset manager, interested in piloting AI-enabled modules
- aย quantitative or technical team, keen to develop models or extensions
- anย academic or AI researcher, looking for deployment bridges
โฆwe invite you to reach out. Join us in building the open, collaborative infrastructure that unlocks next-generation asset management.
Contact us atย [email protected]ย to learn more, discuss collaboration, and see whatโs next.
]]>– ๐๐ฃ๐ง-๐ข๐ฆ๐ฆย is an open-source variant of GPT-4, giving developers full model weights, local deployment options, and freedom to fine-tune. This means enterprises can run AI on-premises, adapt it for niche use cases, and fully control sensitive data.
– ๐๐ฃ๐ง-๐ฑ, released last night, is OpenAIโs most advanced model yet. It delivers stronger reasoning, faster responses, longer context memory, improved coding ability, and more accurate multimodal understanding across text, images, and structured data.
๐ช๐ต๐ ๐ถ๐ ๐บ๐ฎ๐๐๐ฒ๐ฟ๐
ยทย ย The open-source AI race is officially on (Meta, Mistral, xAIโฆ now OpenAI joins in)
ยทย ย ย Organisations can blend GPT-OSS for privacy and control with GPT-5 for cutting-edge intelligence
Atย Finmars, weโve believed in this shift from day one.
Thatโs why our platform isย open-source, modular, and API-first. Ready to integrate models like GPT-OSS to bring intelligence directly to your financial infrastructure. ]]>
The financial sectorโs value added per employee is about double that of any other sector in Luxembourg, generating an average of โฌ236,400 in gross value added in 2024.
Economic impact:
โข Finance accounts for 14% of jobs and about 25% of total economic value added
โข Every job in finance creates 1.1 additional jobs elsewhere in the economy
Where does the value come from?
51.8% โ Banking
30.8% โ Insurance
13.2% โ Fund Industry
4.2% โ Consulting and Legal
Tax and public finance contribution:
The financial sectorโs corporate tax contribution more than doubled from โฌ3.4 billion in 2014 to โฌ7.2 billion in 2024 and it now represents:
~64% of corporate income tax revenue
~80% of wealth tax revenue
Why it matters:
Luxembourgโs finance workers generate higher output per person than any other sectorโmaking them a key driver in a nation that ranks #1 globally for GDP per capita.
As global competition, decentralization, and open-source innovation reshape the industry, how can Luxembourgโs finance sector future-proof itself against global shifts?
Could open, modular financial infrastructure be part of the answer?
]]>๐๐ก๐๐ญ ๐ฆ๐๐ค๐๐ฌ ๐ญ๐ก๐ข๐ฌ ๐๐ข๐๐๐๐ซ๐๐ง๐ญ:
โขย Complete transparency:ย Model weights, source code, and training datasets will be fully open and reproducible. No black boxes.
โขย Apache 2.0 license: Free for everyone to use and adapt: businesses, researchers, governments, individuals. True public good AI.
โขย 1,000+ languages:ย Including underrepresented ones like Quechua and Romansh. This isn’t just another English-centric model.
โขย 100% Swiss hydropower:ย Trained on the “Alps” supercomputer with 10,000+ Nvidia Grace Hopper Superchips. Sustainable AI done right.
๐๐ฐ๐จ ๐ฉ๐จ๐ฐ๐๐ซ๐ก๐จ๐ฎ๐ฌ๐ ๐ฏ๐๐ซ๐ฌ๐ข๐จ๐ง๐ฌ:
โขย 8B parameters for everyday accessibility
โขย 70B parameters competing with top proprietary systems
The Swiss AI Initiative involves 800+ researchers and represents something we rarely see: ethical AI development with strict adherence to Swiss privacy laws and EU AI Act provisions.
๐๐ก๐ ๐๐ข๐ ๐ช๐ฎ๐๐ฌ๐ญ๐ข๐จ๐ง:ย Can openness triumph over corporate ecosystems?
You can find more info about it here: https://ethz.ch/en/news-and-events/eth-news/news/2025/07/a-language-model-built-for-the-public-good.html ]]>
Despite lobbying efforts from over 100 tech firms, including Google and Meta, the EU is proceeding with enforcement.
The first wave began in February 2025, targeting unlawful AI practices and mandating AI literacy for providers.
Whatโs next?
โขย August 2025ย โ General-purpose AI models (including open-source and API-based systems) must meet regulatory requirements.
โขย August 2026ย โ High-risk AI applications in sectors such as healthcare, finance, education, and law enforcement must comply.
This legislation marks a shift in AI governance. It introduces binding obligations not just on AI use, but on leadership accountability, documentation, and oversight.
As AI becomes an operational risk, many firms still lack:
โข Audit trails
โข Governance structures
โข Legal accountability
โข Data transparency ]]>