Arvind Sathi

Deepanker Venkatesh and I recently authored a perspective on the ISDA Common Domain Model (CDM), a foundational development shaping the future of capital markets. ISDA CDM is more than a standard. It represents: • Industry-wide consistency in how data and workflows are defined and executed. • Operational efficiency and risk reduction through streamlined post-trade processes. • Strategic readiness for regulatory alignment and cross-market interoperability. For market participants, the opportunity lies in moving from awareness to action. Engaging early with CDM will better position organizations to capture efficiencies, enable innovation, and strengthen competitive advantage. I welcome thoughts on how CDM will influence the operating model of the industry in the years ahead. Nageswar Cherukupalli, Meena Athinathan, Edward R. Merchant, Anke Bivens, Cary Ellexson, Sathish Ranganathan, Sachin Khatri, Vikesh Gupta, Andreas Pilleris

37

8 Comments

Satisficing Ensembles represent a powerful shift in how we approach complex AI decision-making. Instead of relentlessly pursuing *optimal* solutions – which can lead to instability and errors – this technique focuses on identifying solutions that are ‘good enough’ before refining them. This approach builds robustness into AI systems, reducing catastrophic failures and improving performance under uncertainty. Imagine applying this to financial risk modeling, supply chain optimization, or even AI copilots where consistent, reliable outputs are paramount. It’s about scaling AI with confidence and minimizing costly errors. 🚀💡📈

9

1 Comment

As enterprises scale their digital, analytics, and generative AI initiatives, reliable data is no longer optional it’s foundational. Yet, increasing data consumption, complex multi-cloud ecosystems, and diverse stakeholders are making reliability and cost management more challenging than ever. Data observability provides the visibility and control needed to keep data trustworthy. From monitoring pipeline health and compliance to enabling anomaly detection and automated insights, it’s becoming a strategic enabler of confident decision-making. Everest Group’s latest PEAK Matrix® Assessment on Data Observability Technology Providers explores how the market is evolving with adoption accelerating, buyer priorities maturing, and the value chain expanding to include cost and AI/ML augmentation. Read on: https://okt.to/BOXvzI Get in touch: Chetan Dabade, Samikshya Meher #EverestGroup #DataObservability #DataTrust #Analytics #GenAI #AI #DigitalTransformation #DataReliability #PEAKMatrix #MachineLearning

3

I’m pleased to share that EXL has been granted 10 new U.S. patents for AI-driven innovations spanning insurance, healthcare, retail, utilities, and financial services. These patents reflect how we are making data AI-ready, automating complex workflows, and shaping domain-specific AI that works in the real world - not just in theory.   As Anand Logani, EVP & Chief Data and AI Officer, noted, these patents reinforce our focus on building differentiated #data and #AI capabilities that deliver measurable outcomes for clients.   What matters most to me is what sits behind these patents: deep #domain expertise, disciplined engineering, and talented teams that understand how to turn technology into sustained business impact.

309

7 Comments

How LLM Inference Works by Arpit Bhayani #LLM #inference transforms text prompts into responses through a process involving #tokenization, #transformer layers with self-attention mechanisms, and autoregressive #tokengeneration. There are two stages in practice: the model first handles your full prompt in parallel, then switches to generating tokens one by one, which shifts the bottleneck from math to memory access. Key optimizations include KV caching to avoid redundant computation, batching to improve GPU utilization, and quantization to reduce memory pressure. https://lnkd.in/eJkBiyWE

1

As AI's frontier continues to advance and these systems move into widespread deployment, policymakers, developers, and deployers must answer the question of what meaningful transparency actually looks like in practice. The Hiroshima AI Process reporting framework is among the first serious international efforts to operationalize transparency for AI at the organizational level, but its success ultimately depends on whether it generates information that policymakers and downstream users can actually use. This week at Brookings, we’re convening a group of leading experts from across policy, industry, and civil society to examine early lessons the first cycle of HAIP reporting, where it falls short, and how it can evolve to better support safe, reliable, and accountable AI systems across the entire AI stack. We’re grateful to our collaborators at CDT and honored to feature a keynote from Ambassador Vinay Mohan Kwatra, alongside an outstanding group of expert panelists who are actively shaping AI policy and the development of the technology. Join us in person or online — registration link in Valerie's post below!

2