Nitin Kishore Sai Samala

Claude, ChatGPT, Snowflake Intelligence, Cortex Analyst, Databricks Genie, Zenlytic, TextQL. Think about any AI agent you connect to your data. They all get questions from business users like: “What was the average number of daily active users in the last quarter?” These AI agent connects to Euno’s MCP server to retrieve metadata. That’s when Euno gives these agents the missing context they need to understand your data. ** Just a reminder that Euno is a metadata platform so we only access your metadata ** Euno runs a semantic search and looks for dashboards that use DAU (which is semantically equivalent to Daily Active Users). It filters for certified dashboards using the trust signals you’ve already set in Euno. In this example Euno finds two candidates: 1. DAU used in the User Engagement dashboards 2. DAU_NEW used in the Product Statistics dashboard Both dashboards are certified. But Euno also shows how often each has been used in the past 30 days, along with the owner or creator. With that information, the business user can confidently choose the right one and the AI agent can deliver the correct answer. The same applies to tools like Cursor. Let’s say you’re an engineer and ask: “Can you find me the Tableau dashboard for Customer Success?” Cursor will reach out to Euno’s AI assistant to get that BI context because Cursor is focused on dbt development and doesn’t have access to that layer. You can even go further and ask which dbt models power that dashboard and how those models are calculated. Euno feeds Cursor with that real-time context as you build. It’s not just for business users. It’s for developers too.

25

5 Comments

Semantic Enrichment is rapidly becoming a critical differentiator in LLM performance. It moves beyond simply *feeding* data to models and focuses on augmenting that data with detailed, contextually relevant metadata – essentially, teaching the model *how* to think about the information. This allows for more nuanced reasoning, improved accuracy in complex tasks, and unlocks the potential for truly intelligent automation. Think enhanced decision-making copilots, more reliable risk assessments, and scalable knowledge management systems. 🚀 By investing in semantic enrichment pipelines, organizations can dramatically improve the ROI of their LLM initiatives. 💡📈 #LLMs #AIStrategy #SemanticAI #DataIntelligence #KnowledgeManagement #AIAutomation #Reasoning #EnterpriseAI #GenAI #ToolUse

2

Skills That Matter for Building LLM, RAG, and Agentic AI Applications Building LLM-, RAG-, and agentic AI- applications is less about any single discipline and more about the amalgamation of skills across engineering, data science, and product thinking. Teams achieve this amalgamation not through deliberate pairing, shared ownership of system behavior, and cross-functional reviews. Here’s how I think about the skills that need to come together - and how teams can intentionally blend them. Core skills that cut across LLM, RAG, and agentic systems • problem framing and task decomposition • prompt and instruction design as an interface, not magic • retrieval design: what to fetch, when, and why • evaluation thinking beyond accuracy (failure modes, drift, edge cases) • understanding probabilistic outputs vs deterministic expectations • designing guardrails, constraints, and human-in-the-loop flows These systems succeed or fail at the system level, not the model level. System engineering skills that become critical System engineers already bring strengths in scalability, reliability, and integration. What’s new is learning to operate in a stochastic world. Key shifts: • deterministic → probabilistic behavior • idempotent logic → best-effort reasoning • fixed outputs → distributions and confidence • debugging code → diagnosing behavior What system engineers need to add: • a working understanding of how models learn from data • why LLMs can sound confident and still be wrong • how retrieval, memory, and tools shape outcomes • how to design for uncertainty, not eliminate it You don’t need to become a data scientist - but you do need to understand the underbelly of AI: what’s real signal, what’s pattern matching, and where illusions come from. Data science skills that complement this shift Data scientists bring deep strengths in: • statistical thinking • uncertainty, bias, and variance • model evaluation and validation What many now need to add: • software engineering discipline (APIs, services, observability) • system-level thinking beyond isolated models • product and workflow awareness • operating constraints: latency, cost, reliability, governance Beyond concepts, some concrete skills matter: • designing retrieval pipelines and embeddings • building evaluation harnesses for LLM behavior • understanding orchestration frameworks and tool use • measuring performance at the application level, not just model metrics The real shift LLM, RAG, and agentic applications sit between disciplines. They require: • engineering judgment • statistical thinking • product sense • and systems design The teams that succeed won’t be the ones with the “right titles.” They’ll be the ones that deliberately blend these skills — and know where human judgment still belongs. #AI #EnterpriseAI #LLM #RAG #AgenticAI #AIEngineering #ResponsibleAI #TYS #RayAI #AILiteracy #IntelligenceInBits

9

1 Comment

What is Apache Parquet's Content Defined Chunking (CDC)? Let's zoom into how Parquet is physically laid out. Parquet stores data as row groups -> column chunks -> data pages. These data pages are not just arbitrary splits - they are the unit at which compression and encoding are applied. In many ways, they are the smallest meaningful unit of storage inside a Parquet file. The problem is that Parquet does not define these pages based on the content itself. Page boundaries are determined by size thresholds or row counts. So even if only a small portion of the dataset changes, the system ends up rewriting and re-uploading large parts of the file. And once that happens, compression output changes, byte layout changes, and from a storage perspective, everything looks new. This is why incremental workloads on Parquet behave like full rewrites, and why deduplication at the storage layer becomes ineffective. Content-Defined Chunking (CDC) comes into play here by redefining how these data pages are formed. Instead of cutting pages based on size, CDC uses "rolling hashes" to determine boundaries based on the content itself. The effect of this is: identical sequences of data produce identical page boundaries, regardless of where they appear in the file or how the dataset evolves over time. So, things like compression becomes consistent for unchanged regions, byte layouts remain stable, and only the pages that actually contain modified data need to be rewritten. The rest of the file can remain untouched & more importantly, reusable. BTW, this is already adopted today by companies like Hugging Face for their storage system ("Xet") Hugging Face used to store over 30 PB of models, datasets, and spaces in Git LFS repositories, and that meant any change to a file = re-uploading the entire dataset, which is super expensive. Parquet CDC gives them stable data pages, and Xet deduplicates those across dataset versions - so unchanged data is never re-uploaded. At scale, that directly translates to faster uploads and significantly lower storage overhead. Also amazing to see this now land in Apache Arrow's Rust implementation. (CDC has been around in PyArrow and C++) and there are plans for integrating it in the Apache Iceberg world as well. Learn more from the diagram & comments. #dataengineering #softwareengineering

94

6 Comments

15 Papers That Defined NLP & LLMs (2017–2025) If you want to understand how modern language models evolved — not just what changed, but why — these are the papers to study. Together, they trace how we moved from attention to reasoning in AI. 1. Attention Is All You Need (2017) The paper that started it all. • Introduced the Transformer • Removed recurrence for parallelism • Backbone of GPT, BERT, and more 🔗 https://lnkd.in/gh84Xb-C 2. BERT (2018) Transfer learning redefined NLP. • Pretrain + fine-tune approach • Bidirectional context understanding 🔗 https://lnkd.in/gvCbb2jy 3. GPT-3: Few-Shot Learners (2020) Scale became the new paradigm. • 175B parameters • Prompting as a new interface 🔗 https://lnkd.in/gdn3D6gg 4. T5 (2020) Unified NLP as a text-to-text framework. • One model for multiple tasks 🔗 https://lnkd.in/gDNU-XSF 5. Scaling Laws (2020) Mapped how performance grows with model size. • Predictable scaling curves 🔗 https://lnkd.in/gswH6-3v 6. RAG (2020) Blended retrieval with generation. • Grounded models with external knowledge 🔗 https://lnkd.in/gyu_ZiJy 7. LoRA (2021) Made fine-tuning efficient. • Low-rank adaptation for affordability 🔗 https://lnkd.in/gYREMpEA 8. Chain-of-Thought Prompting (2022) Enabled reasoning step by step. • Better logic and problem solving 🔗 https://lnkd.in/gvwt8TJZ 9. Self-Consistency (2022) Improved reasoning reliability. • Multiple reasoning paths for accuracy 🔗 https://lnkd.in/gG_R2NHa 10. In-Context Learning (2022) Explained how LLMs learn from examples. • Foundation of contextual reasoning 🔗 https://lnkd.in/gm9JCBWy 11. Instruction Tuning (2022) Aligned models with human intent. • Improved conversational behavior 🔗 https://lnkd.in/gNaknD4F 12. Toolformer (2023) Models that learn to use APIs autonomously. • Early step toward planning and tool use 🔗 https://lnkd.in/gMXePE6P 13. ColBERTv2 (2022) Efficient and accurate retrieval. • Scales to billions of documents 🔗 https://lnkd.in/g_N2tT3g 14. LLMs as a Judge (2023) Models evaluating other models. • 85% human-level agreement 🔗 https://lnkd.in/g25MdgT2 15. DeepSeek-R1 (2025) Reasoning through reinforcement learning. • Step-by-step logical thought 🔗 https://lnkd.in/gPHh3URb This isn’t just a reading list — it’s a timeline of progress. Each paper marks a milestone in how AI learned to understand, create, and reason. Which paper do you think changed NLP the most? Share your thoughts below — and help others learn. 📄 Credits: Analytics Vidhya

120

16 Comments