Samaya AI

This might surprise you. We believe that as AI systems advance, human judgment becomes more — not less — important. In her recent conversation with Rana el Kaliouby, Ph.D. on Pioneers of AI podcast, our Founder and CEO Maithra Raghu shares a perspective we don’t hear often enough: As these systems generate more output, the role of humans shifts — from gathering information to providing context, judgment, and decision-making frameworks. That means there could be more need for humans in the loop. At Samaya, we aren’t shooting for AI to be helpful. The output needs to be decision-grade — exhaustive, precise, and auditable — so it can actually be used to take action. To do this we need humans to help with expertise, judgement and intuition. That’s how you move from information to conviction. AI isn’t replacing the work, it’s reshaping how it happens.

Great to see AI Street covering what we’re building at Samaya AI. In the piece, our AI Lead Ashwin Paranjape talks about the challenge of running reliable AI analysis across large financial datasets, and why bigger context windows alone don’t solve the problem. Our Agent Control Plane (ACP) is designed to orchestrate multi-step workflows, verify outputs, and make long-horizon financial analysis reliable at scale. Thanks to Matt Robinson for the thoughtful deep dive. 💡 Read the full article linked in the comments below.

📣 Today we’re excited to announce the launch of Samaya AI’s Agent Control Plane (ACP) and new investment from NVentures (NVIDIA’s Venture Arm) and Databricks Ventures. 📣 Samaya AI’s Agent Control Plane: ACP is a breakthrough architecture for building and running next-generation AI agents that reason in real time with expert-level context for financial services. Why this matters: Generic AI agents struggle with financial complexity - they hit context limits, compound errors, and break down as tools scale. The ACP solves this with purpose-built architecture combining integrated planning, long-horizon execution, advanced reasoning, and optimized context management. Real-world impact: Samaya is supporting workflows from earnings analysis to economy-wide scenario modeling. Through natural language instructions, users can customize agents while maintaining full transparency over how conclusions are reached. This is the blueprint for next-generation AI agents in finance. The investment: We're also thrilled to welcome new investment from NVentures and Databricks Ventures. Working together will support continued development of the ACP and its role as the institutional intelligence layer for financial services. 🤝 #AI #FinancialServices #EnterpriseAI #Agents #Investing

🎉 We are excited to share a big milestone for Samaya: we’ve received our SOC 2® report ✅🔒🎉 This reflects our commitment to strong controls around security, availability, and confidentiality - so customers can confidently use Samaya in their most critical research workflows. Huge thank you to the team and partners who helped make this happen 🙌 #SOC2 #Security #FinTech #AIinFinance

🚀 New Partnership Announcement 🚀 We’re excited to share that Samaya is partnering with Third Bridge Group, bringing trusted expert network insights directly into AI-powered investment research workflows. With this integration, investment professionals can access and synthesize Third Bridge’s expert interviews inside Samaya - combining deep qualitative insight with AI-driven analysis in one unified platform. The result: faster due diligence, richer context, and more differentiated investment theses. We’ve published a short blog and product demo showing how this works in practice 👇 👉 Read more here: https://lnkd.in/gCq9cJ87 #SamayaAI #InvestmentResearch #AIinFinance #ExpertNetworks #FinTech

We’re excited to share SLIM (Simple Lightweight Information Management), a new open-source framework for long-horizon research, built with researchers from Princeton University. Existing systems often get lost in the maze of context, flooding the LLM’s memory with irrelevant content that derails reasoning. SLIM fixes this through: - Decoupled search and browse tools: returning only snippets, not full pages, to minimize noise - Selective content extraction: fetching only the most relevant sections - Periodic summarization of trajectories: compressing long searches while preserving clarity Across BrowseComp and Humanity’s Last Exam (HLE) benchmarks, SLIM outperforms existing frameworks while using only 15–25% of the tool calls and reducing hallucination by 2.5x (47% to 19%). A simpler approach to context management can make deep, multi-step research both cheaper and more accurate.  🔗 Blog, paper, and code in comments. Howard Yen Ashwin Paranjape Yuhao Zhang Thejas Venkatesh Jack Hessel @DanqiChen

We are thrilled to see our CEO, Maithra Raghu, recognized for her visionary work driving Samaya’s expert AI Agents. Congratulations! 🎉 🚀 #ET40UnderForty #LeadershipInAI #SamayaAI

📣 We’re thrilled to announce that Samaya AI will be attending the Stanford Computer Forum Career Fair on October 8, 2025! 📣 Ready to build the next generation of fully autonomous AI agents? At Samaya, we're building AI that collaborates with experts. We innovate across research, engineering, and design. 🚀 We have deep roots in Stanford NLP, with Christopher Manning as an investor, and several Stanford alums on our team who are excited to share what we’re building - and to hear about your work too! 🤝 Come chat about our roles, our vision, and how you can be part of shaping the future of knowledge work with us. See you there! 💙 #StanfordUniversity #Careers #AIResearch #AIAgents #Hiring

📣 Employee Spotlight: Meet Vadim Nikulin 📣 Vadim has over 20 years of experience working on high-loaded distributed systems, with the past decade focused on building machine learning products. After years at Microsoft and Facebook, he made the switch to startups, drawn by the faster pace and bigger impact of building from scratch. At Samaya, one of Vadim’s proudest projects has been designing and scaling our data sources system — a foundation that enables us to manage vastly larger and more complex datasets. In just a year, the volume of stored data has grown by nearly 100x, requiring constant innovation to balance efficiency, precision, and cost. He describes Samaya’s culture a 'surprisingly mature' for a startup: robust tools and processes (SOC 2 compliance, structured APIs, Kubernetes, Terraform), yet still with the agility and energy of an early-stage company. Vadim thrives on the cross-functional nature of his role, collaborating with colleagues across the UK and US. He values both the structured online rituals and the spontaneous in-person conversations that spark ideas and build momentum. Asked why someone should join Samaya, Vadim says simply: 👉 “It’s a great place to learn and build useful products.” Outside of work, Vadim is just as inventive. He raises chickens 🐔, welds quirky contraptions like sidecars for bicycles 🚲, and collects driving tests as a hobby 🚗. We’re lucky to have Vadim’s expertise, curiosity, and creativity on the team! 🌟

Big news at Samaya: our Founder and CEO, Maithra Raghu, has been named to the TIME100 AI 2025 list! 🎉 Through her AI research at Google Brain, working with the "Godfather of AI" Geoffrey Hinton, to building Samaya AI, Maithra has been driven by the vision of Expert AIs that can partner in discovering key insights and high-stakes decision making. We’re so proud to see her recognized for the vision that drives what we’re building at Samaya. Congratulations, Maithra!