Trang Nguyen

“The Ontology is the substrate through which humans and AI work together to create operational value — rooted in your business, compliant with your norms, and measured against the outcomes you care about.“ — Palantir Chief Architect, Akshay Krishnaswamy 🎤 ⬇️

571

12 Comments

No A/B test? No problem. Learn how Regression Discontinuity Design (RDD) can uncover causal effects in observational data by exploiting sharp cutoffs. Dive into the theory and a real-world e-commerce example in Alejandro Alvarez Pérez's newest article.

21

2 Comments

"GenAI interpretability gets less attention than the constant deluge of model releases, but it is arguably more important." I completely agree with this point from the recent Urgency of Interpretability blog by Dario Amodei. As foundation models evolve rapidly, the lack of interpretability can become a major barrier to safe, compliant, and large-scale adoption—especially in finance. To build trust, we must be able to explain model outputs clearly and reliably. I’m excited to speak on this topic at the upcoming AI4 2025 conference, where I’ll present our latest work: “Beyond the Black Box: LLM Interpretability in Finance.” I’ll be sharing how we’re getting into the "brain" of LLMs to uncover meaningful financial features and apply them across key use cases like trading and sentiment analysis. Thankful to my co-author, Ariye Shater, for his collaboration on this work. Links: - Paper Link: https://lnkd.in/exrmXh97 - Conference: https://ai4.io/finance/ - Dario's Blog: https://lnkd.in/ePDp-x-W #GenAI #Interpretability #Finance #LLMs #AI4Conference #SparseAutoencoders #Anthropic #ResponsibleAI #MechanisticInterpretability

87

High-performance AI is not built in a vacuum but on strong foundational engineering. I spent the holiday break completing my 7th Databricks certification: Machine Learning Associate. While it is tempting to stay at a strategic 30k foot view, I believe an AI executive must stay grounded in the technical reality their teams face every day. In large-scale corporate environments, architecture is where the time dimension becomes critical. A shortcut taken today might deliver a quick win, but the right architectural choices ensure that you are building a scalable asset for the next five years rather than technical debt that paralyzes the organization in eighteen months. Why I believe technical rigor is a business imperative: 🔹 Precision in Execution: I was reminded this week how a simple shift from classic Python UDFs to Pandas UDFs on Spark can transform performance. It sounds like a detail, but it can be the difference between a sluggish pipeline and a cost-optimized engine that potentially saves 100k Euro in annual cloud spend. 🔹 Trust and Reliability: In a recent production use case, we achieved >92% prediction accuracy and cut processing time from hours to seconds. You do not reach that scale by accident. You reach it by validating production workflows against the latest platform capabilities. 🔹Empowering the Team: As a leader, staying current is not about doing the work for your team. It is about setting the standard. Practical experience with frameworks like ISO 27001 and 42001 provides the safety rails for high-speed delivery instead of bureaucratic roadblocks. I have found that the most effective technical leadership ensures engineering precision translates eventually into documented business impact. The objective is rarely just to build a model but to architect the industrial-grade platforms that makes those solutions sustainable and reliable for the long term. I am curious: What foundations are you laying today to ensure your Data & AI strategy scales effectively over the next years?

78

16 Comments

If your datasets are poorly labeled, you’re inviting a host of issues like noise, bias, and inconsistencies that can seriously hurt your model's accuracy and reliability when it’s out in the wild. That's why data labeling has evolved from being just a preliminary step into a vital part of engineering and operations in the AI lifecycle.

7

New spec to bridge semantic definitions between AI and BI. And a great list of partners in the mix.

1

Scientific knowledge is expanding faster than ever but the data behind it remains fragmented. Genes, diseases, drugs, publications, datasets… all living in separate silos. Nextnet Ontology brings these pieces together into a unified semantic framework that connects biological entities, relationships, and evidence across the life sciences landscape. The result: 🔸 More meaningful discovery 🔸 Smarter AI reasoning 🔸 A deeper understanding of how knowledge truly connects Read the full blog to see how ontology powers the foundation of the Nextnet platform 👉 https://lnkd.in/ebr3VE_a

161

Universities would be wise to reform their curriculum to focus on data-oriented design and functional product-based outcomes, rather than standardized testing, to better prepare the workforce. https://lnkd.in/gacD9_BF

3

2 Comments

"The next wave of AI success is going to be completely dependent on structured data... Data is the mature, logical, boring middle, but if you want an honest assessment of where all the smart AI is going, look to those mature, logical, boring leaders pushing palatable hypergrowth." https://lnkd.in/gprqvzpz

9

1 Comment

The strongest data warehouses succeed because modeling, governance, and shared meaning guide every design decision. When metadata and standards drive the process, teams trust their data and scale without confusion. ER/Studio gives organizations a model-driven foundation that keeps definitions consistent, integrations clean, and communication clear across technical and business teams. Benefits to the business: 💡 Faster approvals and smoother delivery ✔️ Consistent KPIs and trusted analytics 🔹 A scalable framework that adapts as platforms evolve Read the blog 🔗 https://lnkd.in/eR-xsFsE #ERStudio #DataArchitecture #DataGovernance #KnowledgeGraph #DataProducts

11

1 Comment

Warehouses were built for humans running reports—not agents firing evolving queries at ms cadence. We break down stateful scaling limits, manual tuning traps, and how agent-native engines close the gap (with MCP + StarRocks). Dive into the details: https://hubs.la/Q03DL_7D0 #Lakehouse #DataInfrastructure #DataAnalytics #DataEngineering

11

The semantic layer and context engineering are rapidly becoming essential differentiators in Data and AI.   Why is this important? A robust semantic model—essentially a way to define relationships between nodes and edges based on ontology—prevents AI from hallucinating and ensures it delivers trustworthy, explainable, and repeatable insights.   AI cannot be solely blamed for inconsistencies. If terms like “revenue” or “active customer” vary across businesses, it is crucial to communicate this to AI to eliminate ambiguity and drive consistent answers and insights.   On one hand, semantic modeling structures data and provides contextual understanding based on meaning. On the other hand, context engineering adds situational awareness, ensuring that semantic interpretations align with current scenarios, resulting in more nuanced and intelligent behavior.   Semantic modeling is no longer a mere “data governance nice-to-have”; it is the foundation for scaling AI initiatives.   In the landscape of Snowflake, Databricks, and the Open Semantic Interchange (OSI), we see different approaches:   - Snowflake embeds semantic definitions within the database (semantic views), aiming for consistency but potentially locking users into its ecosystem. - The OSI initiative promotes interoperability across platforms, a vital vision still in its early stages and reliant on vendor collaboration. - Databricks adopts a more future-proof strategy by building semantics on open formats, Unity Catalog, and metadata governance, making them portable, governable, and AI-ready across multi-cloud and multi-tool environments.   Databricks is pulling ahead with its focus on agentic AI enablement. Through Mosaic LLM Serving Endpoints, agent bricks, and automation layers connected to the semantic model, Databricks transforms semantics from mere definitions into actionable intelligence.   This translates to: - LLM agents capable of reasoning over trusted business terms - Seamless orchestration of AI-driven data quality and automation workflows - Portability and governance without sacrificing flexibility   My POV: Semantic modeling is the key success enabler for enterprise AI. Without it, AI outcome can be hit and miss. With it, we unlock trustworthy, explainable, and automatable decisions.   What’s your take — will enterprises bet on platform-native semantics or the open, agent-ready approach Databricks is pushing? 😊 #AgenticAI #Databricks

20

From the front line of delivering Agentic systems Forget the pilot and the launch party. The real game of agentic AI starts when it goes live. In 2026, a lot of Agentic systems are shifting from isolated pilot programs to full enterprise-scale deployment. Being in the middle of this shift, I see these areas to focus on 1. Token Discipline. This is the key to control cost for Agentic systems. Treating tokens like a precious currency and implementing real-time metrics for every agent request. Without this, costs spiral and the ROI vanishes. Tokens are the measure of Cost Of Goods Sold (for those of you in finance LOL). 2. Think agent Observatory, not dashboards. Develop or integrate with systems that evaluate agent accuracy, detect subtle failures, and trigger automated playbooks is top of mind. I see observability as the smoke detector for customer trust. Without one we are flying blind, and the first one who will notice that our system doesn’t deliver value is…our customer. 3. Institutionalize Lineage. This often overlooked. Every data source, prompt version, model iteration, reasoning result and evidence should be traceable. Version everything. And in all this, my observation is also that the definition of the secret sauce, our intellectual property, and competitive advantage, has shifted from traditional AI models a few years ago to the agentic reasoning prompts and the data layers and knowledge graphs powering these agents. Interesting times! #AIStrategy #AgenticAI #LiveAgents 2026

22

Cloud data platform Snowflake's principal data strategist Jennifer Daniell Belissent, PhD tells Verdict that asking how to win the AI Space Race is the wrong question. 'Coopetition' and an open source approach is the way forward, says Belissent. #ai #airegulation #opensource #coopetition https://lnkd.in/eeHqhzNE

10

2 Comments

#LLm #rag Combining LLM fine-tuning with Retrieval-Augmented Generation (RAG) offers superior performance compared to using either technique alone. Fine-tuning allows the LLM to specialize in a particular domain, while RAG provides real-time access to up-to-date information, leading to more accurate and contextually relevant responses. This hybrid approach overcomes the limitations of relying solely on either fine-tuning or RAG. Here's why the combination is better: - Enhanced Accuracy and Domain Expertise: Fine-tuning improves the LLM's ability to understand and generate responses within a specific domain, while RAG ensures the LLM has access to the most current information available. - Mitigation of Hallucinations: By grounding the LLM's responses in retrieved data, RAG significantly reduces the risk of hallucinations - Adaptability to Dynamic Information: RAG enables the LLM to incorporate real-time updates and new information. - Handling Complex Queries: The combination of fine-tuning and RAG allows the LLM to tackle complex questions that require both in-depth domain knowledge and access to specific facts. - Improved User Experience: The enhanced accuracy, relevance, and up-to-date information provided by the combined approach leads to a more satisfying user experience.

16

🎙️ It was an absolute pleasure to join Kyle Winterbottom on the Driven by Data Podcast for a deep dive into the role data, analytics & AI is playing in Brenntag’s transformation journey. We’re not just talking about data, we are reshaping decision-making by unlocking measurable business value through the power of Data & AI. From building “purple teams” to driving EBITA impact, this episode explores the real-world complexities of scaling value at enterprise level. At Brenntag, partnerships, talent and AI are more than enablers, they’re catalysts. Grateful to share our story and the lessons we’re learning along the way.

23

2 Comments

AI systems often blend data and inference making it harder to trace bias back to its source. Orion is built with intention. By separating data pipelines from inference, it anchors decisions in verified customer systems, not generated assumptions. That structure gives us greater control over bias and keeps insights aligned with reality.

6

When evaluating AI-native security platforms, the most important question is how conclusions are formed. Many tools layer AI on top of fragmented data, which limits outputs to summaries without real context. Without an ontology-driven reasoning layer, AI cannot connect assets, identities, permissions, and activity into a coherent understanding of risk. Architecture determines whether AI produces explainable decisions or confident noise. Read more: https://bit.ly/49GniUF #AISecurity #SecurityArchitecture #CISO #CloudSecurity #SecOps

12

1 Comment

Executive confidence in AI outcomes is not keeping pace with executive spending on AI tools. That gap is not a technology problem. It is an organizational readiness problem. New platforms get purchased. Pilots get launched. Yet six months later, processors still default to manual workflows because no one built the foundational habits required to make the technology stick. The tools sit idle while the invoices keep coming. This pattern repeats across industries because leadership often confuses procurement with transformation. The flawed assumption here is that AI adoption is a deployment event rather than a capability-building process. Buying software does not change operational DNA. Building literacy does. When a workforce lacks the baseline fluency to evaluate AI outputs, question its limitations, and integrate it into daily decision-making, even the best tools become expensive shelf-ware. Before approving the next AI investment, ask a harder question: does your team have the habits to absorb it? #AIStrategy #OperationalExcellence #AIAdoption #ExecutiveLeadership #ThePhoenixFirm

3