Domyn

📣 AI evolves, as do we. Meet Domyn – the new name behind the company making AI yours to own, govern, and trust. As our mission grows, so does our identity, reflecting the focus and commitment we place on building sovereign AI that drives growth, protects intellectual property, and aligns with the responsibilities of mission-critical sectors. Stay tuned - more information to come! Read more in our press releases: 🇺🇲 bit.ly/43GUEzv 🇮🇹 bit.ly/3FLIltJ

🗓️ This April, our CEO and Founder @uljansharka will join global leaders at SEMAFOR World Economy — the leading modern platform for global CEOs and executives — to sit down with Semafor editors for a conversation on the most critical issues facing the new world economy. ➡️ We hope to see you there: semaforworldeconomy.com

📊 LLMs are increasingly used to extract structured knowledge from complex financial documents, but how can trust what they produce? 🔎 In their latest research, Fabrizio Dimino, Abhinav Arun, Bhaskarjit Sarmah, and Stefano Pasquali introduce FinReflectKG – EvalBench, the first systematic evaluation framework for financial Knowledge Graph extraction from real-world SEC 10-K filings. By benchmarking multiple extraction strategies across dimensions like faithfulness, precision, and relevance, the team uncover critical trade-offs that single metrics miss. If you’re working on financial AI, compliance, or knowledge extraction, this work offers a foundation for more transparent and reliable evaluation. ➡️ Read the full blog to learn more: shorturl.at/JQVx2

The conversation around future-proofing success with AI is gaining momentum. Our CEO and Founder @uljansharka was interviewed by @Founders_Forum as part of a survey of 200+ top leaders on AI readiness for leadership success. Featured in an exclusive report, the findings confirm one of our core beliefs at Domyn: the future of successful AI adoption won’t be driven by performance alone, but also by solutions that are reliable, trustworthy, and responsible in real-world deployment. Get the full framework and discover how to become AI-ready for long-term success: shorturl.at/sxA0C

LLMs are gaining momentum in the finance sector, powering everything from investment research to credit risk analysis. Many expect an “off-the-shelf” model to be neutral enough to deliver unbiased output. Yet evidence suggests that this assumption is far from accurate. In a research paper co-authored by Fabrizio Dimino, Bhaskarjit Sarmah, Krati Saxena, and Stefano Pasquali, we demonstrated that examined representation bias in Qwen LLMs—showing how biases introduced as early as the pre-training stage can distort outputs and ultimately influence financial decisions. Explore the dynamics of these biases in the paper, and discover why finance companies must rely on domain-specific, finance-tuned LLMs to ensure reliability, mitigate structural skews, and safeguard decision-making quality. ➡️ shorturl.at/dyvXY

🌍 Incredible momentum for Africa’s AI community! The AI Hub for Sustainable Development has unveiled the first cohorts of innovators who will help shape the continent’s AI future. 🔎 This connected ecosystem links long-term AI infrastructure development with immediate access to compute resources and training for startups, researchers, and public institutions across Africa. ➡️ As one of their technology partners, we’re proud to help drive this initiative forward by equipping innovators with the tools they need to create long-term impact.

🚀A huge thank you to everyone at AI Dev 25 for hosting such an energizing and insightful event. It was a pleasure to be part of it!

Stefano Pasquali, @DomynAI's Head of Financial Services, at AI Dev 25 x NYC: "Domyn believes in building transparent, auditable and sovereign AI ecosystems, where innovation meets accountability."

📊 In financial AI, complex questions rarely have simple answers. They require connecting data points across years, companies, and filings — a process known as multi-hop reasoning. Yet most LLMs struggle with this because evidence is often scattered across unstructured text. That’s why our researchers Abhinav Arun, Reetu Raj Harsh, Bhaskarjit Sarmah, and Stefano Pasquali developed FinReflectKG–MultiHop, a new benchmark that uses structured, knowledge-graph evidence to enable more efficient and interpretable reasoning and deliver better, more accurate answers. For AI-powered financial analysis, this new development is a crucial step forward, pushing beyond surface-level reading to connect the dots across time, risk, and performance — ultimately helping systems reason more like analysts. ➡️ Read the full paper to learn more: shorturl.at/dRGPw

Ever wondered what goes on behind the scenes when models “learn to learn”? ➡️ The ICL Visualizer take you there, with a backstage look at how Large Language Models learn from context in real time. Unveiled by our AI engineer Reetu Raj Harsh at the @VISxAI conference, this interactive platform shows models detecting patterns, processing examples, and making predictions before your eyes. 🔎 Step backstage and watch AI in action: shorturl.at/zgJQR

Financial decision-making is not a guessing game — yet most financial analysis tools stop at spotting correlations without explaining why things happen, leaving professionals in the dark about the causes behind the numbers and offering little insights to make fully informed decisions. All that’s about to change. Our research team — Bhaskarjit Sarmah, Stefano Pasquali, Abhinav Arun, Alejandro Zuñiga — is advancing reliable AI-powered financial analysis with FinCARE: a hybrid framework that combines causal discovery algorithms, financial Knowledge Graphs from SEC filings, and the reasoning power of LLMs. The result? Not just smarter insights, but real predictive power—AI that doesn’t just crunch numbers, but reveals the causes behind them. In short, an approach that gives financial institutions full awareness and confidence to make decisions with clarity and foresight. Discover more: shorturl.at/KZcGK

Our CEO and Founder @uljansharka will participate in two panel discussions during #FII9 by @FIIKSA in Riyadh. Come join him in the following sessions: 1. Can AI Make Markets Perfectly Efficient? 📍 Pavilion 2 – LAB 4 🗓 October 28th | ⏰ PM – PM 2. App-Layer Futures: Generalization or Specialization — Which Path Dominates? 📍 Pavilion 2 – LAB 4 🗓 October 30th | ⏰ AM – AM From discussing trustworthy domain-specific LLMs for regulated industries to shedding light on how leveraging AI can enhance asset value in real time, we look forward to connecting and exchanging ideas with peers and innovators alike!

Our partnership with @Allianz is expanding. Five years ago, we set out to make conversational AI available to enterprises. Today, after successfully completing our AI pilot, that vision is scaling. What’s coming up next for this collaboration? Find out in our news article: shorturl.at/0JtjL

🔔 Our SVP and Head of Financial Services Solutions, Stefano Pasquali, will be speaking at #NeurIPS2025. 🗓️ When: Saturday, December 6th 📍 Where: San Diego Convention Center Bringing together experts from AI, finance, and regulation, the #NeurIPS2025 workshop will explore how to adapt generative AI to the unique demands of the finance and risk management industries, enabling the responsible development of solutions for real-world scenarios. Learn more about the event: shorturl.at/qRR8U

In AI, every natural-language question hides a complex challenge: translating everyday language into precise database queries. In finance, this task is especially difficult due to complex databases and highly specialized terminology, and because small errors can lead to costly inaccuracies and regulatory risks. Yet until now, there hasn’t been a dedicated benchmark to properly train or evaluate AI models on this critical task within the financial domain. Enter FINCH — the first-ever dataset and evaluation framework designed specifically for AI systems that answer questions using real financial data. Developed by Bhaskarjit Sarmah, Stefano Pasquali, and Avinash Kumar Singh, the project lays the foundation for building AI that truly understands finance — not just language. Their new research paper, “FINCH: Financial Intelligence using Natural language for Contextualized SQL Handling,” dives deeper into this breakthrough, opening the door to more reliable and domain-aware AI solutions for financial services. Take a look now: shorturl.at/D3o2i

🚀 The Domyn Platform is here. The agentic AI platform built for the world’s most regulated work has just dropped — empowering you to build, configure, orchestrate, and govern proprietary AI agents with full control. Here’s what the platform brings to your organization: ✅ Integrated intelligence: To design agents that unify language, logic, and workflows by connecting your enterprise systems to LLMs through the platform ✅ LLM compatibility: To connect with Domyn’s LLM series or integrate with any compatible large language model of your choice ✅ Intuitive, no-code configuration: To empower teams to build and customize agents even without technical expertise ✅ Frictionless integration: To embed the platform into your systems in a fast and easy way thanks to a dedicated SDK ✅ Private deployment: To gain full control over monitoring, scaling, and provisioning while ensuring complete privacy by deploying in a private, isolated area ✅ Personalized options: To create custom agents or choose from Domyn’s catalog of specialized and customized agents. The Domyn Platform has already been adopted by a subset of selected customers. Discover more: domyn.com