Dhruv Singh

Awesome post by the Databricks team https://lnkd.in/gGQ9ptRe My summary: They trained a model called KARL that beats Claude 4.6 and GPT 5.2 on enterprise knowledge tasks (searching docs, cross-referencing info, answering questions over internal data), at ~33% lower cost and ~47% lower latency. The key insight: instead of throwing expensive frontier models at enterprise search, you can use reinforcement learning on synthetic data to train a smaller model that's faster, cheaper, AND better at the specific task. RL went beyond making the model more accurate. I t learned to search more efficiently (fewer wasted queries, better knowing when to stop searching and commit to an answer). They're opening this RL pipeline to Databricks customers so they can build their own custom RL-optimized agents for high-volume workloads. I think we'll continue to see data platforms become agent platforms. Databricks' KARL paper is really an agent platform play. The pitch: you already store your enterprise data in the Lakehouse, now Databricks will train a custom RL agent that searches and reasons over it, tuned specifically for your highest-volume workloads (workloads = apps = agents). The business move is closing the loop: data storage → retrieval → custom agent training → serving, all on Databricks. They're turning "your data lives here" into "your agents live here too." Kudos Ali Ghodsi Reynold Xin Matei Zaharia

54

1 Comment

New release from Packt Publishing… “Building Business-Ready Generative #AI Systems — Build Human-Centered Generative AI Systems with Context-Aware Agents, Memory, and LLMs for the Enterprise” See it at https://amzn.to/3Jdcio5 𝓚𝓮𝔂 𝓕𝓮𝓪𝓽𝓾𝓻𝓮𝓼: 🔵Build an adaptive, context-aware AI controller with advanced memory strategies 🟢Enhance GenAISys with multi-domain, multimodal reasoning capabilities and Chain of Thought (CoT) 🟠Seamlessly integrate cutting-edge OpenAI and DeepSeek models as you see fit

14

3 Comments

𝗧𝗶𝗽𝗮𝗹𝘁𝗶 + 𝗡𝗲𝘁𝗦𝘂𝗶𝘁𝗲: 𝗔𝗣 𝗔𝘂𝘁𝗼𝗺𝗮𝘁𝗶𝗼𝗻 𝗧𝗵𝗮𝘁 𝗦𝗰𝗮𝗹𝗲𝘀, 𝗦𝗲𝗰𝘂𝗿𝗲𝘀, 𝗮𝗻𝗱 𝗦𝗶𝗺𝗽𝗹𝗶𝗳𝗶𝗲𝘀 𝗙𝗶𝗻𝗮𝗻𝗰𝗲 What’s the biggest challenge your AP team faces today—and what tools are you using to solve it? From Boston’s biotech startups to Philly’s SaaS firms and DC’s nonprofits, finance teams across the Northeast Corridor are upgrading their A𝗰𝗰𝗼𝘂𝗻𝘁𝘀 𝗣𝗮𝘆𝗮𝗯𝗹𝗲 (𝗔𝗣) 𝘄𝗼𝗿𝗸𝗳𝗹𝗼𝘄𝘀 to stay agile, compliant, and fraud-resistant. A typical 𝗔𝗣 𝘄𝗼𝗿𝗸𝗳𝗹𝗼𝘄 includes: - Invoice intake - PO matching - Approval routing - Payment execution - Reconciliation and audit When managed manually, this process is slow, error-prone, and vulnerable to fraud—especially across multiple entities or international vendors. Tipalti automates the entire AP lifecycle. It uses AI and OCR to digitize invoices, match them to POs, route approvals, and execute global payments. It supports multi-entity operations and integrates seamlessly with NetSuite, syncing financial data in real time and eliminating manual entry. But automation isn’t just about speed—it’s also about security. Tipalti’s Detect® fraud prevention module uses machine learning to flag suspicious patterns—like duplicate invoices, shared payment details, and unusual vendor behavior. It blocks high-risk payees, tracks fraud attempts, and maintains detailed audit trails for compliance. Why is this critical? Because 𝗶𝗻𝘃𝗼𝗶𝗰𝗲 𝗳𝗿𝗮𝘂𝗱—where scammers submit fake or altered invoices to trick businesses into unauthorized payments—is one of the fastest-growing threats in finance. Tipalti helps stop fraud before it hits your books. The *𝗧𝗶𝗽𝗮𝗹𝘁𝗶 + 𝗡𝗲𝘁𝗦𝘂𝗶𝘁𝗲 𝗶𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗶𝗼𝗻 delivers: ✅ Real-time financial sync ✅ Touchless invoice processing ✅ Supplier self-service portals ✅ Automated tax and regulatory compliance ✅ Up to 80% reduction in AP workload At 𝗨𝗻𝗶𝘃𝗲𝗿𝘀𝗮𝗹 𝗘𝗾𝘂𝗮𝘁𝗶𝗼𝗻𝘀, we help Northeast finance teams implement intelligent AP automation tailored to their growth goals. Whether you're scaling across NYC, Newark, Baltimore, or DC—we’re ready to help. #AccountsPayable #Tipalti #NetSuiteIntegration #UniversalEquations

1

OpenAI-backed Chai Discovery just raised a $130M Series B at a $1.3B valuation—and it’s a clean snapshot of where AI + biotech is heading. Chai is building foundation models tuned for drug discovery, essentially a “computer‑aided design suite for molecules.” Their latest model, Chai 2, is already showing materially better success rates in de novo antibody design—designing new antibodies from scratch, not just tweaking existing ones. That’s the kind of shift that can compress timelines and failure rates across the entire discovery stack. The round was led by General Catalyst and Oak HC/FT, with participation from Menlo Ventures, OpenAI, Dimension, Thrive, Neo, Yosemite, SV Angel and others—bringing total funding north of $225M. This isn’t “AI wrapper” money; this is a bet that model-native biotech becomes core infrastructure for pharma and advanced therapeutics. If platforms like Chai actually deliver CAD-for-molecules at scale, how do you see it reshaping who captures value in drug development—incumbent pharma, model platforms, or full-stack new entrants? https://lnkd.in/eGpjSh3E #AIinBiotech #DrugDiscoveryPlatforms #BioVC

6

1 Comment

Databricks and Perplexity co-founder Andy Konwinski is forming the Laude Institute, a new AI research institute, and backing it with $100 million of his own money. The Laude Institute will function less as a traditional research lab and more like a fund making investments structured as grants. It will focus on "Slingshots" (early-stage research) and "Moonshots" (long-horizon labs tackling species-level challenges). The institute's board includes AI luminaries like UC Berkeley's Dave Patterson, Google's Jeff Dean, and Meta's Joelle Pineau. The institute announced its first flagship grant of $3 million annually for five years to anchor the new AI Systems Lab at UC Berkeley. The new fund aims to catalyze work that doesn't just push the field of AI forward, but guides it toward more beneficial outcomes. Congratulations to Andy Konwinski on this incredible commitment to AI research. TechCrunch: https://lnkd.in/dy7x9zAT #AI #Research #AIforGood #Philanthropy #VenturePhilanthropy #Databricks #Perplexity #UCBerkeley

2

Another breakthrough in #AI was published earlier today by Sapient Intelligence, and the linked headline by VentureBeat says it all. https://lnkd.in/giYJtEb9 TL;DR, the new #HierarchicalReasoningModel (#HRM) reasons more like a human by quickly thinking about the best way to solve a problem and then thinking more slowly to solve the problem. This new method can be trained on an order of magnitude fewer training examples and can solve multi-step puzzles that simply befuddle #CoT #LLM s, like #ChatGPT and #Claude, ultimately requiring far fewer resources.

6

2 Comments

Microsoft's market cap declined far more in one day than all of the capital they have invested in LLMs and related infrastructure to date. Of course the market cap is in a bubble due to the LLM hype but for investors that bought yesterday it's still an 11% loss in one day. Those who have never experienced a market crash should be careful. It's very unpleasant, and can be catastrophic for those who are over-leveraged. https://lnkd.in/gKaV8hyy

7

2 Comments

A strong strategy is only effective when paired with flawless execution. 🎯 The Codingscape team - now part of AROBS Group achieved this by implementing Airtable as a unified execution layer, replacing 95% of unstructured planning within six weeks.⚡ Context matters. This project was for Roblox, an organization with 2,500 employees, over 100 million daily active users, and more than $4 billion in revenue. At this scale, aligning strategy with execution is complex. The solution: * A unified, AI-ready execution layer built on Airtable. * Full visibility from C-level initiatives to daily operations. * Teams continued using their existing tools, such as Jira and Asana, while leadership gained real-time, comprehensive visibility. The results: * Strategy is visible across all layers of the organization. * Clear alignment between frontline execution and organizational priorities. * Streamlined cross-team requests and enhanced capacity planning. Delivered 2 weeks ahead of schedule by a 5-person team. This is what senior execution looks like: speed, precision, and zero noise. Congratulations to the Codingscape team, led by Porter Haney. 👏 I am proud of the team and look forward to achieving more with Airtable. This is impact, done right., done right. This is part of a broader AROBS strategy to help companies leverage AI and low-code / no-code tools such as Airtable to implement solutions faster, more efficiently, and with greater impact. More details about the project here https://lnkd.in/dJVEnUfY #DigitalTransformation #EnterpriseScale #ProductLeadership #Airtable #Codingscape #AROBS

60

4 Comments

Databricks now lets you run OpenAI models (like GPT-5) directly on your enterprise data. Most AI tools or products force you to move your data into their product — adding cost, risk, and endless API hops. This changes that. Now, agents inside Databricks can securely work with large, complex business datasets — accelerating AI-driven processes and giving you faster answers to your most critical questions. At LBMC we remain committed to building AI readiness through data readiness in Databricks. #AI #Databricks #EnterpriseAI #Agents https://lnkd.in/eUYefPhR

21

1 Comment

Without IBM, I'm not sure if Apache Spark would've succeeded in the way it did. Even though Hadoop came out as the first mover ahead of Spark, Spark was able to become the dominant open source project with the backing of major enterprise leaders like IBM. IBM's contributions to code and marketing shifted Spark from an academic project out of Berkeley-- One that came out 4 years behind Hadoop -- into a trusted enterprise technology. I think it's going to be the same for AI memory. MemMachine (fka Intelligent Memory) is coming out into the open source and we're looking for the next IBM to work with. Don't miss out.

8

Salesforce just admitted "autonomous" AI agents don't work. Last week, CIO.com reported Salesforce is adding deterministic scripting layers to Agentforce because LLM-based autonomy keeps failing in production. Their internal term for it? "Doom-prompting"—engineers stuck rewriting prompts endlessly because agents drift, hallucinate, and give different answers to identical queries. The platform pitched as "self-directed agents that resolve issues end-to-end" now needs workflow mapping, data modeling, and rule-based guardrails. Sound familiar? It's the same story as 11x earlier this year. Here's what nobody wants to say: You can't prompt-engineer your way out of a data architecture problem. The agents that actually work in production aren't "autonomous." They're: → Grounded in structured data with clear provenance → Orchestrated through explicit handoffs, not "let the LLM figure it out" → Feeding engagement signals back into the data model The winners in Agentic GTM won't have the smartest agents. They'll have the best data underneath.

74

2 Comments

Context engineering shifts LLMs from oracle to analyst—frame the task, curate tokens, and apply reusable patterns (RAG, tool calls, memory) to keep complex systems robust. A must-read for product and Agile teams building trustworthy AI-enabled workflows. Explore Chris Loy's take: https://buff.ly/Ca2GrrO #Product

1

Starburst has always believed in providing customers with the freedom of choice. Nowhere is this more true than the emerging developments around interoperable compute among data platforms. Recently, I sat down with Snowflake’s Ryan C. Green to talk about the Open Semantic Interchange (OSI), an open, vendor-neutral standard for sharing semantic models across AI, BI, and analytics tools. OSI gives customers a common, open way to define and share business metrics. This helps them stay consistent across dashboards, notebooks, and machine learning models, no matter which tools they use. For Starburst, this furthers one of our core goals. Choice. Customers should be able to keep one trusted set of business definitions inside their Snowflake environment while still having the freedom to choose the tools that work best for them. OSI helps make that possible. It marks an important step in bringing Starburst’s engine to Snowflake-interoperable compute. The video from my chat with Ryan is coming soon 👀 In the meantime, you can get more details from our recent press release: https://lnkd.in/ev2XGB4u #OpenSemanticInterchange #OSI #Interoperability #AI #Data #Snowflake #Starburst

182

8 Comments

Oracle was the first database I worked on, back in 2006, at BT Group, alongside my pair-programming buddy, Ritesh Agarwal, and our tech lead, Aditya S.. Along with SQL Server, Oracle forms the operational core for many enterprises, and for decades, people had to choose between them or migrate across them painstakingly. Interestingly, it's no longer about PL/SQL vs T-SQL; natural language is the new query language, allowing Tursio to support both SQL Server and Oracle! Language is no longer a barrier, and data can stay exactly where it is. #sqlserver #oracle #tursioai #fridaystory

23

Last night reminded me why I started Philly Data & AI. You put the right people in a room and something shifts. Practitioners, builders, people who are tired of talking about AI and ready to actually do something with it. That was last night. Philadelphia has no shortage of people leading this charge. Tim Dodd, Tempest Carter, and so many others in this community are doing the real grassroots work: convening people, driving cross-sector collaboration, and pushing toward something bigger than any one organization. This is what civic AI leadership actually looks like. Tonight is the follow-up. Tempest’s Tech Talk on pilots isn’t abstract theory. It’s the part most people skip: how do you take a promising idea and actually validate it before committing everything? That’s how we build responsibly and at scale. Together we are making Philadelphia a model for what ethical, community-driven AI innovation can look like. That doesn’t happen at the top. It happens in rooms like last night’s. I’m planning to be there tonight. Come find me if you are too. #PhillyAI #CivicTech #EthicalAI #AIInnovation #Philadelphia #PhillyTech #CommunityFirst

16

Databricks stock is not happening, as the custom database development company wants to stay private. Databricks sales topped a $4.8 billion revenue run-rate during the third quarter and is growing 55% year-over-year. With its latest VC capital infusion, the company plans major investments in AI.

1