<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Dr. Michael Gerstlauer on Medium]]></title>
        <description><![CDATA[Stories by Dr. Michael Gerstlauer on Medium]]></description>
        <link>https://medium.com/@michael.gerstlauer?source=rss-8950c36d8332------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*l2B2lboq3EkQGuT6</url>
            <title>Stories by Dr. Michael Gerstlauer on Medium</title>
            <link>https://medium.com/@michael.gerstlauer?source=rss-8950c36d8332------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 05:47:46 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@michael.gerstlauer/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[From Reactive RCA to Prescriptive Excellence: How Snowflake Rewires Manufacturing Quality]]></title>
            <link>https://medium.com/@michael.gerstlauer/from-reactive-rca-to-prescriptive-excellence-how-snowflake-rewires-manufacturing-quality-d1cb728ba096?source=rss-8950c36d8332------2</link>
            <guid isPermaLink="false">https://medium.com/p/d1cb728ba096</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[quality-management]]></category>
            <category><![CDATA[snowflake-cortex]]></category>
            <category><![CDATA[ishikawa]]></category>
            <category><![CDATA[snowflake]]></category>
            <dc:creator><![CDATA[Dr. Michael Gerstlauer]]></dc:creator>
            <pubDate>Sat, 28 Feb 2026 15:12:28 GMT</pubDate>
            <atom:updated>2026-02-28T15:24:39.087Z</atom:updated>
            <content:encoded><![CDATA[<p>In the high-stakes world of manufacturing, <strong>quality management</strong> has historically been a reactive, rearview-mirror exercise. A defect appears, a line goes down, and suddenly everyone is in a room staring at dashboards, spreadsheets, and PDF manuals, trying to reverse‑engineer what happened.</p><p>We don’t lack tools; we lack <strong>intelligence in the middle</strong> of the operation — a connected brain that understands products, processes, and context in real time.</p><p>With the <strong>Snowflake AI Data Cloud</strong>, manufacturers are starting to move from predicting failures to <strong>prescribing excellence</strong>. By putting AI directly into the center of the shop floor data flows, Snowflake turns classic quality techniques — <strong>5 Whys</strong>, <strong>Ishikawa diagrams</strong>, <strong>Gemba</strong>, and <strong>Six Sigma (DMAIC)</strong> — into continuous, AI‑driven workflows instead of one‑off workshops.</p><p>Very quickly, <strong>two killer apps</strong> emerge in this journey:</p><ul><li><strong>Snowflake Intelligence for root‑cause analysis</strong> — conversational 5 Whys and automated Ishikawa, powered by the full history of your operations, not just what people remember in a meeting.</li><li><strong>Cortex Code for sensor‑to‑insight automation</strong> — taking new sensor and event streams and turning them into production‑ready pipelines, features, and dashboards in minutes.</li></ul><p>Below is a practical blueprint for how you can use these capabilities and accelerators to eliminate unplanned events and take control of your production line.</p><h3>1. Conversational “5 Whys” and Automated Ishikawa — Powered by Snowflake Intelligence</h3><p>The <strong>5 Whys</strong> and <strong>Ishikawa (Fishbone)</strong> are not new ideas. What <em>is</em> new is what happens when you let <strong>Snowflake Intelligence</strong> run them end‑to‑end on top of your entire manufacturing data estate.</p><p>Instead of a group of engineers guessing in front of a whiteboard, you get an <strong>AI quality co‑pilot</strong> that:</p><ul><li>Understands your products, lines, and assets because it sits on top of your <strong>unified Snowflake model</strong> (ERP, MES, historian, quality, maintenance)</li><li>Has full access to <strong>historical patterns</strong>, not just what people remember from the last incident</li><li>Can turn a “Why?” question into a <strong>multi‑step reasoning chain</strong> across all relevant signals in seconds</li></ul><p>This is why Snowflake Intelligence becomes the <strong>killer app</strong> for RCA in manufacturing.</p><h3>Snowflake Intelligence as a 5 Whys Engine</h3><p>When a line goes down today, most teams lose hours just assembling data:</p><ul><li>Structured data from <strong>ERP/MES</strong> (orders, work centers, shifts, operators)</li><li>Time‑series data from <strong>SCADA/PLC and historians</strong></li><li>Unstructured content like <strong>PDF work instructions, maintenance manuals, and inspection reports</strong></li></ul><p>With Snowflake, all of that sits in <strong>one governed platform</strong>. Snowflake Intelligence (via <strong>Cortex Analyst, Notebooks, or Native Apps</strong>) can then drive a <strong>conversational 5 Whys</strong> loop over this shared truth.</p><p>Example:</p><ol><li>You ask: <em>“Why did OEE drop on Line 2 yesterday?”</em></li></ol><ul><li>Snowflake Intelligence queries your OEE model and responds:</li><li>Availability loss from 14:07–14:32</li><li>Caused by an unplanned stop on Filler 3</li><li>Linked to Order 4711 and SKU ABC‑123</li></ul><p>2. You ask: <em>“Why did Filler 3 stop?”</em></p><ul><li>It correlates alarm logs and sensor data: repeated <strong>“torque too high”</strong> events on the capper head, rising vibration over 5 days, plus a missed lubrication task in the CMMS.</li></ul><p>3. You ask: <em>“Why was lubrication missed?”</em></p><ul><li>It checks maintenance records and shift plans and finds a staffing gap and an overloaded technician route on the late shift.</li></ul><p>The classic 5 Whys becomes a <strong>guided reasoning session</strong> where Snowflake Intelligence constantly:</p><ul><li>Joins the right tables and time windows</li><li>Surfaces the next most likely factor</li><li>Keeps a <strong>transparent trace</strong> of each “Why” and the underlying evidence</li></ul><p>You still make the final call — but the heavy lifting of data orchestration and hypothesis testing is done by the AI.</p><h3>Automated Ishikawa Over the “6 Ms”</h3><p>The same engine that powers conversational 5 Whys can also <strong>auto‑generate and quantify an Ishikawa diagram</strong>.</p><p>Instead of a blank fishbone, you give Snowflake Intelligence a clear question:</p><ul><li><em>“For Product X on Line 4, over the last 3 months, break down all defects by the 6 Ms (Man, Machine, Method, Material, Measurement, Environment) and rank each branch by impact on scrap and rework cost.”</em></li></ul><p>Because the data already lives in Snowflake, the AI can:</p><ul><li>Map events and defects to the <strong>6 Ms</strong> using your semantic model and historical patterns</li><li>Quantify each branch using <strong>FPY loss, scrap cost, or downtime minutes</strong></li><li>Highlight specific combinations (e.g., <em>Material from Supplier Y + Night Shift + Line 4 Capper</em>) that systematically drive defects</li></ul><p>The result is not just a prettier fishbone. It’s a <strong>living Ishikawa</strong>, backed by:</p><ul><li>Real data, not memory</li><li>Automatic updates as new orders, defects, and maintenance events land in Snowflake</li><li>Direct links into downstream actions (e.g., open a CAPA, adjust inspection plans, trigger additional checks for specific lots)</li></ul><p>In other words, Snowflake Intelligence turns Ishikawa and 5 Whys from <strong>one‑time workshops</strong> into a <strong>continuous RCA service</strong> that runs on top of your manufacturing data cloud — and that is the real killer app for modern quality management.</p><h3>2. Bringing Data to the Gemba</h3><p><strong>Gemba (現場)</strong> means <em>“the real place”</em> — where value is actually created. Traditionally, a Gemba walk means going to the line, observing the process, and talking to operators. But observation without data context quickly turns into <strong>opinion theater</strong>.</p><p>Snowflake changes Gemba by making <strong>live, contextual data available in the moment</strong>, on the floor.</p><h3>Live Comparison in Your Pocket</h3><p>Because all your key manufacturing data (ERP, MES, historian, quality, maintenance, supply chain) lives in <strong>one governed platform</strong>, you can expose <strong>mobile‑friendly views or apps</strong> directly on Snowflake:</p><ul><li>During a Gemba walk, a supervisor can ask:</li><li><em>“How is this line performing vs. last month?”</em></li><li><em>“Show me FPY for this product and shift over the last 7 days.”</em></li><li><em>“Are today’s micro‑stops on this filler normal compared to last week?”</em></li></ul><p>Live dashboards and conversational UI — powered by Snowflake and Cortex — allow you to anchor each observation in <strong>facts instead of assumptions</strong>.</p><h3>Andon in Near Real Time</h3><p>The traditional <strong>Andon</strong> approach waits for a physical signal — a light, a button, a call for help — after an issue already impacts the process.</p><p>With Snowflake, you can:</p><ul><li>Continuously ingest telemetry and events via <strong>Snowpipe Streaming</strong> or <strong>Kafka/IoT integrations</strong></li><li>Define <strong>centralized quality rules and control limits</strong> as SQL models and dynamic tables</li><li>Use <strong>Cortex</strong> to watch for patterns in near real time</li></ul><p>Instead of waiting for a tower light to flash, your “digital Andon” in Snowflake:</p><ul><li>Detects early drifts in <strong>SPC charts, defect rates, or process capability</strong></li><li>Pushes alerts into <strong>Teams/Slack</strong> or operator UIs</li><li>Suggests <strong>immediate corrective actions</strong> (e.g., parameter tweaks, additional checks, temporary holds)</li></ul><p>Gemba becomes a <strong>closed‑loop system</strong> where what you see on the floor is continuously backed by what your data sees in the cloud.</p><h3>3. Six Sigma DMAIC at Snowflake Speed</h3><p><strong>Six Sigma’s DMAIC</strong> (Define, Measure, Analyze, Improve, Control) is still the backbone of structured quality improvement. In reality, most programs stall in the <strong>Measure</strong> and <strong>Analyze</strong> phases because data is hard to access and models are painful to maintain.</p><p>Snowflake removes this friction by making DMAIC <strong>data‑native</strong>.</p><h3>Define &amp; Measure: Baselines in Seconds</h3><p>Instead of spending weeks to get a clean baseline, your quality team can:</p><ul><li>Use curated models on Snowflake to immediately answer:</li><li><em>“Which product–line–shift combinations created the highest scrap and rework costs over the last 12 months?”</em></li><li><em>“What does FPY look like for Product Family A by plant, line, and supplier?”</em></li><li>Pull <strong>12‑month FPY trend lines</strong> directly from Snowflake in seconds and slice by:</li><li>SKU, plant, line, shift, operator group</li><li>Supplier, material lot</li><li>Tooling and maintenance events</li></ul><p>Your <strong>“Measure”</strong> phase becomes a SQL query and a quick visualization, not a 6‑week data project.</p><h3>Analyze: Semantic Models, Not Manual Plumbing</h3><p>Traditionally, Black Belts and data engineers have to jointly build dimensional models and cube‑like structures before any analysis can start. With Snowflake and AI, you can:</p><ul><li>Let an <strong>AI assistant infer semantic models</strong> from your existing:</li><li>Dashboards and their queries</li><li>Frequently used joins and filters</li><li>Naming conventions (product, asset, shift, defect categories)</li><li>Automatically propose:</li><li>Business entities (Product, Line, Shift, Operator, Supplier, Defect Type)</li><li>Relationships and grain</li><li>KPIs (FPY, OEE components, scrap rate, rework cost, MTBF, etc.)</li></ul><p>Instead of months of manual modeling, you get a <strong>semantically rich layer in minutes</strong>, which both humans and AI agents can reason over. Your teams can then focus on <strong>“Improve”</strong> and <strong>“Control”</strong>, not on plumbing.</p><h3>Accelerators That Make This Real</h3><p>At this point, two <strong>killer apps</strong> emerge for quality on Snowflake:</p><ol><li><strong>Snowflake Intelligence for 5 Whys &amp; Ishikawa</strong> (your continuous RCA service)</li><li><strong>Cortex Code for Sensor-to-Insight Automation</strong> (turning new signals into production-grade assets in minutes)</li></ol><p>Together, they shorten the cycle from <em>“something went wrong”</em> to <em>“we understand it and have adjusted the process”</em>.</p><h3>Cortex Code: From Sensor to Quality Dashboard in 3 Minutes (Second Killer App)</h3><p>One of the biggest killers of continuous improvement is the lag between <strong>adding a new sensor</strong> and <strong>seeing it in a trusted dashboard or model</strong>.</p><p>With <strong>Snowflake Cortex Code</strong>, process and quality engineers — technically savvy people without formal coding backgrounds — can describe what they want in natural language and let AI scaffold the technical work. Unlike generic coding assistants or low‑code platforms, Cortex Code is <strong>Snowflake‑native</strong>: it understands your warehouses, schemas, policies, and manufacturing data model, so it can safely generate artifacts that actually run in your environment and can be adapted as the shop floor changes.</p><p>A typical interaction looks like this:</p><ul><li><strong>Input:</strong></li><li>A new data source (e.g., an MQTT/Kafka stream or historian table already landing in Snowflake)</li><li>A short prompt like:</li></ul><pre>Create a quality monitoring pipeline for torque and temperature on Line 2 Filler. - Join to work orders, SKUs, and shifts from the MES view FACT_PRODUCTION. - Calculate 1‑minute rolling averages and standard deviations per asset. - Flag points above 3 sigma for use in control charts.</pre><p><strong>Cortex Code then generates, for example:</strong></p><ul><li>Snowflake objects for <strong>ingestion and storage</strong> (tables, stages, streams)</li><li><strong>Snowpark / SQL code</strong> for cleaning, aggregating, and feature‑engineering signals, such as:</li></ul><pre>create or replace table QUALITY_FEATURES_LINE2 as<br>select<br>  t.event_ts,<br>  t.asset_id,<br>  f.work_order_id,<br>  f.sku,<br>  avg(t.torque_nm) over (<br>    partition by t.asset_id<br>    order by t.event_ts<br>    range between interval &#39;1 minute&#39; preceding and current row<br>  ) as torque_avg_1m,<br>  stddev_samp(t.torque_nm) over (<br>    partition by t.asset_id<br>    order by t.event_ts<br>    range between interval &#39;1 minute&#39; preceding and current row<br>  ) as torque_std_1m,<br>  case when abs(t.torque_nm - avg(t.torque_nm) over (<br>                partition by t.asset_id<br>                order by t.event_ts<br>                range between interval &#39;1 minute&#39; preceding and current row<br>              )) &gt; 3 * stddev_samp(t.torque_nm) over (<br>                partition by t.asset_id<br>                order by t.event_ts<br>                range between interval &#39;1 minute&#39; preceding and current row<br>              )<br>       then 1 else 0 end as torque_out_of_control<br>from RAW_TELEMETRY_LINE2 t<br>join FACT_PRODUCTION f<br>  on t.work_order_id = f.work_order_id;</pre><ul><li><strong>Tasks / dynamic tables</strong> to keep everything up to date in near real time</li><li>A starter <strong>dashboard or notebook</strong> that already joins telemetry to orders, assets, shifts, and defects</li></ul><p>From there, your engineers can:</p><ul><li>Plug the new features directly into <strong>control charts and alarm rules</strong></li><li>Feed them into <strong>Snowflake Intelligence</strong> for richer 5 Whys and Ishikawa reasoning</li><li>Re‑generate or extend code as requirements evolve, without starting from scratch each time</li></ul><p>This makes Cortex Code the <strong>second killer app</strong>: every new sensor or data source becomes usable for quality <strong>within minutes</strong>, not sprints.</p><h3>Prescriptive “What‑If” Simulations on Top of the Same Model</h3><p>Once you have:</p><ul><li>A <strong>unified manufacturing model</strong> (orders, lines, shifts, inventory, cost, quality) in Snowflake, and</li><li>Fast iteration on new signals via <strong>Cortex Code</strong></li></ul><p>…you can start to run serious <strong>prescriptive simulations</strong> directly in the AI Data Cloud.</p><p>Typical scenarios:</p><ul><li><em>“What happens to inventory, service level, and scrap if I move SKU X from Plant A Line 1 to Plant B Line 3 next month?”</em></li><li><em>“Which combination of safety stock, inspection frequency, and test limits keeps us within budget while holding FPY above 98%?”</em></li><li><em>“How much quality risk do I add if I relax this test by 5% for a single supplier and shift?”</em></li></ul><p>A practical implementation on Snowflake looks like this:</p><ul><li>Use <strong>parameter tables</strong> (scenarios, levers, constraints) to define the knobs you want to turn</li><li>Implement simulation logic in <strong>SQL and Snowpark</strong> (e.g., recomputing material flows, lead times, and defect probabilities under different settings)</li><li>Let <strong>Cortex</strong> expose this as a conversational interface: <em>“Run scenario S3 vs. S1 and summarize the impact on scrap cost, OTIF, and capacity utilization.”</em></li><li>Store results back into Snowflake for <strong>comparison over time</strong> and for reuse in future RCA sessions</li></ul><p>The same Snowflake Intelligence that explains yesterday’s issue (via 5 Whys and Ishikawa) can now also <strong>play forward</strong>: testing your ideas for change <em>before</em> you touch the line.</p><p>You move from <strong>reaction to rehearsal</strong>. You are not just learning from yesterday’s defects; you are <strong>simulating tomorrow’s risks and opportunities</strong> before they ever reach the shop floor.</p><h3>From Rearview Mirror to Digital Copilot for Quality</h3><p>The shift underway in manufacturing quality is not about yet another dashboard. It’s about making Snowflake the <strong>intelligent middle layer</strong> of your operations:</p><ul><li>Every classic method — 5 Whys, Ishikawa, Gemba, DMAIC — becomes <strong>always‑on and data‑driven</strong></li><li>Engineers get a <strong>conversational copilot</strong> that can see across ERP, MES, historians, quality systems, and maintenance</li><li>Managers go to the <strong>Gemba with facts</strong>, not assumptions</li><li>Six Sigma programs spend their cycles on <strong>improvement</strong>, not data preparation</li></ul><p>By moving intelligence to the center of your operations with the Snowflake AI Data Cloud, you are no longer just reacting to quality issues — you are <strong>simulating the future and prescribing excellence</strong> so that many of those issues never occur in the first place.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d1cb728ba096" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Industrial Data Operations: Unlocking
the AI Factory with HighByte, Novotek and the Snowflake Data…]]></title>
            <link>https://medium.com/@michael.gerstlauer/industrial-data-operations-unlocking-the-ai-factory-with-highbyte-novotek-and-the-snowflake-data-a7915270886f?source=rss-8950c36d8332------2</link>
            <guid isPermaLink="false">https://medium.com/p/a7915270886f</guid>
            <dc:creator><![CDATA[Dr. Michael Gerstlauer]]></dc:creator>
            <pubDate>Tue, 16 Dec 2025 10:38:41 GMT</pubDate>
            <atom:updated>2025-12-16T10:38:41.224Z</atom:updated>
            <content:encoded><![CDATA[<h3>Industrial Data Operations: Unlocking<br>the AI Factory with HighByte, Novotek and the Snowflake Data Cloud</h3><h3>Industrial Data Operations: Unlocking the AI Factory with HighByte, Novotek and the Snowflake Data Cloud</h3><p>Manufacturers today are data-rich but insight-poor. Up to 90% of industrial sensor data goes unused because it lacks the necessary context for IT and line-of-business applications. The fragmentation between Information Technology (IT) and Operational Technology (OT) systems prevents the activation of this latent value, hindering the scalable deployment of AI and advanced analytics for Industry 4.0.</p><p><strong>Industrial Data Operations (DataOps)</strong> is the strategic framework that solves this challenge. It is the operationalization layer that integrates all data — from the shop floor to the enterprise — into a unified system.</p><p>The joint solution of <strong>HighByte Intelligence Hub</strong>, <strong>Novotek</strong> and the <strong>Snowflake AI Data Cloud for Manufacturing</strong> delivers this vision:</p><ol><li><strong>HighByte</strong> provides the Industrial DataOps engine at the Edge, connecting proprietary OT sources, modelling data into a unified, operational context, and standardizing it for the Cloud.</li><li><strong>Snowflake</strong> acts as the unified, scalable AI Data Cloud, consolidating the newly contextualized industrial data with existing enterprise data for AI, ML, and democratized analytics.</li><li><strong>Novotek </strong>with almost 40 years of unparalleled experience in industrial IT and automation, acts as a vital reseller and consulting partner. Novotek provides the deep expertise required to design, implement, and support these complex digital transformation initiatives, ensuring optimal outcomes and long-term success for manufacturers.</li></ol><p>The result is a single, governed, and real-time data foundation that accelerates decision-making, enables predictive maintenance, and drives AI-powered innovation across the global enterprise.</p><h3>The Unified Solution Stack: Edge-to-Cloud Data Activation</h3><h3>1. HighByte: Contextualization at the Edge (The Universal Translator)</h3><p>HighByte Intelligence Hub is the critical <strong>Industrial DataOps software</strong> purpose-built to model and orchestrate data directly at the Edge. It solves the fundamental problem of data fragmentation and lack of context, transforming raw sensor values into meaningful, business-ready information before it leaves the factory floor.</p><p><strong>HighByte’s Core Values</strong></p><ul><li>Securely connects diverse industrial protocols (OPC UA, Modbus, MQTT, etc.) via a script-free interface, eliminating complex, custom code.</li><li>Enriches raw telemetry with vital operational context, such as <strong>Machine ID, Operational Mode, and Timestamp</strong>. This converts a temperature reading from “noise” into an actionable insight (e.g., “CNC Mill 3 is 15°C above normal”).</li><li>Models and structures data into a <strong>Unified Namespace (UNS)</strong> architecture, ensuring data is delivered to consuming applications — including Snowflake — in a consistent, usable format.</li></ul><h3>2. Snowflake: The AI Data Cloud for Scale (The Analytics Engine)</h3><p>Once contextualized by HighByte, the industrial data flows into the Snowflake AI Data Cloud, where it is consolidated with enterprise data from various IT systems. Snowflake provides the single, governed platform needed to truly bridge the IT/OT divide.</p><ul><li><strong>Advanced Analytics &amp; AI:</strong> Provides a scalable, flexible environment to run predictive maintenance models, perform real-time operational analytics, and train sophisticated AI/ML applications.</li><li><strong>Data Democratization:</strong> Breaks down data silos, making clean, standardized data accessible to data scientists, analysts, and business users across the organization.</li><li><strong>Secure Collaboration:</strong> Enables frictionless, live data sharing across departments, vendors, and partners, all underpinned by a robust, built-in governance framework.</li></ul><h3>SAP, Snowflake, and HighByte: The Synergy for IT/OT Integration</h3><p>The recently announced <strong>strategic partnership between SAP and Snowflake</strong> delivers the final piece required for a complete, AI-ready data fabric for manufacturing. This collaboration provides a direct, highly governed bridge between the OT-side context and the business-side context:</p><p><strong>The Challenge:</strong> HighByte adds <em>operational</em> context (What is the machine doing?). But to act on it, you need the <em>business</em> context (What is the Order ID, the Bill of Materials, or the Inventory level?) — which resides in SAP (ERP/MES) systems.</p><p><strong>The Solution:</strong> The SAP/Snowflake partnership, enabled by the new <strong>SAP Snowflake solution extension</strong>, allows for <strong>zero-copy data sharing</strong> between the SAP Business Data Cloud (BDC) and the Snowflake AI Data Cloud.</p><p>This is a game-changer for IT/OT integration:</p><p><strong>HighByte</strong></p><p><strong>Contextualizes OT Data:</strong> Connects to PLCs/SCADA and tags the raw telemetry with <strong>operational details</strong> OT data also comes from time series databases (historians), transactional systems (MES), and files. It’s this mix of real-time and historical data that makes HighByte unique and not just another protocol translation tool.</p><p><strong>SAP BDC</strong></p><p><strong>Provides Business Context:</strong> Houses the <strong>semantically rich data</strong> (Order IDs, Product Specs, Inventory).</p><p><strong>Snowflake</strong></p><p><strong>Unifies the Context:</strong> Receives the contextualized OT data from HighByte <em>and</em> seamlessly accesses the rich SAP business data via <strong>zero-copy sharing</strong>.</p><p><strong>Novotek</strong></p><p><strong>Experts in the OT environment and in Industrial DataOps. </strong>Partner with knowledge about bridging the OT / IT Gap. Being a vital player in the eco-system in bringing seamless solutions to the market.</p><h3>The Combined Outcome: AI Grounded in Business Truth</h3><p>By combining HighByte’s ability to model and standardize OT data with the SAP/Snowflake zero-copy link for business data, manufacturers can now:</p><ul><li><strong>Run Real-Time Predictive Maintenance:</strong> Instantly correlate a machine’s high-vibration reading (from HighByte) with the specific <strong>Warrantee Status, Cost of Goods, and Production Schedule</strong> (from SAP via Snowflake) to immediately calculate the financial risk of downtime and prioritize maintenance.</li><li><strong>Simplify AI/ML Development:</strong> Build AI models directly in Snowflake that are trained on a harmonized dataset combining high-volume sensor data and trusted SAP business data, avoiding complex, manual data duplication and integration.</li><li><strong>Achieve Complete Data Lineage:</strong> Ensure end-to-end data governance and traceability from the raw sensor to the final financial report, enabling faster compliance and root-cause analysis.</li></ul><p>This creates a true, end-to-end data fabric where OT, IT, and ERP systems are unified in one governed platform, accelerating time-to-value for Enterprise AI initiatives.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a7915270886f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Smart Manufacturing: Bridging 2 gaps at the same time: IT/OT and Edge/Cloud Manufacturing]]></title>
            <link>https://medium.com/@michael.gerstlauer/smart-manufacturing-bridging-2-gaps-at-the-same-time-it-ot-and-edge-cloud-manufacturing-77925231efb4?source=rss-8950c36d8332------2</link>
            <guid isPermaLink="false">https://medium.com/p/77925231efb4</guid>
            <category><![CDATA[industrial-data]]></category>
            <category><![CDATA[cloud-edge]]></category>
            <dc:creator><![CDATA[Dr. Michael Gerstlauer]]></dc:creator>
            <pubDate>Wed, 02 Jul 2025 10:56:44 GMT</pubDate>
            <atom:updated>2025-07-02T10:56:44.657Z</atom:updated>
            <content:encoded><![CDATA[<p>Smart manufacturing is a concept that is revolutionizing the industrial landscape and creates a transformative shift towards creating “smarter” factories that are not only more efficient and flexible but also accurate and responsive to customer demand. Snowflake is partnering with Nokia to offer an innovative solution to converge IT/OT as well as Edge and Cloud.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*NIA9AHcGVi_ncZnB" /></figure><p><em>Smart Factories unify shop-floor operational data (OT) with enterprise data (IT), creating a single information flow for real-time insights across the entire process.</em></p><p>To enable smart manufacturing, enterprises need the ability to collect, analyse, and leverage vast amounts of data, which is instrumental in achieving efficient production and quality improvements.</p><p><strong>However, the transition to smart manufacturing is not without its challenges</strong>. Many companies grapple with their legacy technology stack, which is often not designed to handle the scale and speed that the modern business environment demands. Incorporating new data sources often involves long delays and significant investments in hardware and software. The use of data silos to incorporate new data sources or service multiple business stakeholders often results in fragmented data, making it difficult to obtain a holistic view of customers and serve them effectively across various channels, products, or lines of business.</p><p><strong>Moreover, legacy solutions for data sharing, such as flat file uploads or other static transfers, often hold businesses back. </strong>Teams are burdened with managing and maintaining legacy pipelines that are extremely fragile and inflexible, discouraging new partnerships due to the technical challenges they pose. Data governance, security, and maintenance pose additional challenges, often leading to data silos by organisation and process areas.</p><p><strong>In the end-to-end manufacturing process, manufacturers deal with thousands of types of assets, machines, sensors, and meters, each with different logic controls and protocols.</strong> These machines generate data and communicate with each other in hundreds of different device protocols. The traditional separation of organizations and the types of data and systems they support has led to manufacturers spending most of their time acquiring data rather than deriving value from it.</p><p>Moreover enterprises have to think of ways to serve the decision making on the frontline — in Edge — which is described as Industrial DataOps, including Edge DataOps to make data consumable, immediately when produced by sensor, asset or process, for event-driven use cases like digital twins or edge AI.</p><h3><strong>Industrial DataOps</strong></h3><p>Today, enterprises have a plethora of ways to serve decision-making on the frontline. Common for these initiatives is to implement data-driven use cases to increase safety and security, productivity and efficiency. What’s important, ultimately everything circulates around how to effectively consume, process, assess and operationalize relevant data for decision-making tools like AI and digital twins.</p><p>That’s where <strong>edge computing and Edge DataOps </strong>excel by supplementing AI &amp; Data Platforms like Snowflake, enabling edge actively preparing data for edge and cloud-based use cases and simultaneously serving as a managed execution environment for low-latency edge AI, analytics and code execution.</p><p>There are several reasons enterprises are considering bringing advanced decision making capabilities to the frontline, for example extreme low latency requirements, increased reliability, data consumability, and reduced bandwidth and storage cost.</p><p>Before transferring data to a central location (like most already does), contextualization, structure and metadata should be added to data already on edge — to support frontline workers with data-driven decision-making, and to seamlessly serve any related use cases on edge (distributed, shop-floor decision making) or in cloud (centralized, enterprise-wide decision making).</p><p>In other words, edge can support Enterprise Architecture modernization for the age of AI, being the place to generate data products, to support both paradigms of Data Fabric (central governance, distributed execution) and Data Mesh (domain specific, distributed data products owned by domain teams of subject matter experts). With that approach, Industrial Knowledge Graph (KG) to ground Enterprise data for Agentic AI, becomes easier to generate — not to mention that most advanced scenarios enable automated KG updates directly from production assets!</p><p>In addition to that Edge DataOps make data consumable for edge AI, it makes standards-based Industry 4.0 approach implementation simpler, by supporting vertical integration with readily applicable data (data products) and serving horizontal integration (digital thread with Asset Administration Shell (AAS)) with operational data from any production asset — supporting asset lifecycle management with AAS and supply chain-wide carbon emission tracking and reduction with Digital Product Passport (DPP).</p><h3><strong>Manufacturing Data Cloud</strong></h3><p>The Snowflake platform simplifies and streamlines the convergence of IT and OT data systems, enabling Smart Manufacturing use cases and powering high value business outcomes — from yield &amp; quality, to increased output, to lower production costs and more. This secure and economical environment brings out the value in one of your key assets, the data.</p><p>The inclusion of powerful analytics engines enables streamlined <strong>root cause analytics</strong> to address opportunities for improvement while <strong>Predictive and prescriptive analytics</strong> are used to prevent future yield and quality issues. Digitisation across the manufacturing processes means more data driving analytics that enable better energy usage and <strong>equipment uptime</strong>.<br>As analytics improve manufacturing process efficiency and product yield, manufacturers enjoy <strong>increased product output</strong> while avoiding new capital investments in additional equipment.</p><p>The secure integration of enterprise and manufacturing operations data (IT/OT) eliminates silos and enables smart manufacturing use cases.</p><h3><strong>Closing the Loop — Smart Manufacturing</strong></h3><p>Successful introduction of new real-time data-driven capabilities to organization requires seamless connectivity-enabled information continuum. Actionable insights for humans and machine with ecosystem-wide understanding and decisions. Data capabilities must benefit shop-floor employees, enterprise data governance, data scientists and the whole value chain.</p><p>Industrial AI as a hybrid edge-cloud architecture combining for example GenAI, Causal AI, Analytical AI and Agentic AI to achieve industrial autonomy via multiple use cases optimizing overall productivity, safety, compliance and varying operational constraints. Especially on the shop-floor, the variety of algorithms is vast due to the diversity of use cases. ​</p><p>With modern edge computing approaches, industrial operations can achieve central promises of Industry 4.0 — agility via modular solutioning, streamlined maintainability via reduced tech debt and simplified data management via OT asset and protocol abstraction. Edge DataOps brings information and decision-making simplicity on top of data complexity.</p><h3><strong>Edge and Cloud with Nokia MXIE</strong></h3><p><strong>Nokia MXIE as on-prem edge platform</strong> delivers versatile edge-cloud computing and multiple shop-floor connectivity options, with simple deployment of value-adding applications like Crosser for Edge DataOps (and Unified Namespace), and seamless connection to Snowflake Manufacturing Cloud capabilities. This ecosystem approach opens simple and easily maintained, modular way to deploy data-driven enterprise use cases whether they require low latency, reliability or easily scalable computing power of cloud. With Snowflake ML and AI capabilities and MLOps, both edge and cloud industrial AI use cases are well supported and available. MXIE already powers hundreds of private wireless networks and critical communication applications, and is ready to bring clarity to industrial data operations too.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*lWG76zJdqpKD7XUb" /></figure><p><em>Solution Overview of Snowflake, Nokia MXIE and Crosser</em></p><p>Get ready for digital transformation by starting with your most urgent use case, then take advantage of the scalability of compound Snowflake, Crosser and Nokia MXIE capabilities by adding new use cases while needed!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=77925231efb4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why UNS matters for factory connectivity and how Snowflake supports]]></title>
            <link>https://medium.com/@michael.gerstlauer/why-uns-matters-for-factory-connectivity-and-how-snowflake-supports-0cc656e70a69?source=rss-8950c36d8332------2</link>
            <guid isPermaLink="false">https://medium.com/p/0cc656e70a69</guid>
            <category><![CDATA[iiot]]></category>
            <category><![CDATA[unified-name-space]]></category>
            <category><![CDATA[it-ot-convergence]]></category>
            <category><![CDATA[sensor-data-analysis]]></category>
            <dc:creator><![CDATA[Dr. Michael Gerstlauer]]></dc:creator>
            <pubDate>Tue, 18 Feb 2025 14:55:46 GMT</pubDate>
            <atom:updated>2025-02-18T14:55:46.032Z</atom:updated>
            <content:encoded><![CDATA[<p>“Unified Namespace” is a term widely used in the manufacturing community in the context of IT/OT integration or IIOT (industrial IoT), yet it often requires explanation. Furthermore, it is not always natively (or “naturally”) available from data platform providers. Yet is might be one of the most important technological concept that connects IT and OT.</p><p>It all starts with people. If you build a solution blueprint, one has to work with both sides of the house ; IT people and shopfloor people, both experts in there field, but Frequently “translation” is needed between “Factory OT people” and “IT people”. In IT, we talk about “the art of possible”, “fail fast” , “sandboxing”……these are all terms that might create certain headache for a person responsible for a large factory infrastructure, where failure is not an option and trying stuff out is potentially extremely costly. So how to bridge that gap? Let’s start with some basic concepts.</p><p><strong>Factory System Hierarchy</strong></p><p>Well, think of factory equipment as a <strong>“hierarchical pyramid”</strong>, communicating from layer to layer (and of course, there is a norm describing that pyramid, ISO-95).</p><p>It starts with the physical process (what actually happens, Level 0), then field level devices (actuators &amp; sensors that sense and impact the process directly, Level 1), followed by automation and control systems with real time capabilities (e.g. PLCs — programmable logical controllers, Level 2), supervised by Level 3, which are usually systems that look at time frames like days or shifts and oversees operations, workflows and dependencies between single systems (e.g. SCADA systems or MES systems). Finally, level 4 is the “ERP” level (Enterprise Resource planning), which looks holistically at the optimization of the overall operations, often beyond the manufacturing process itself. While this is highly simplified, this represenation gives a reasonable overview over the ISO-95 hierarchy.</p><p><strong>Why is this a challenge for IT?</strong></p><p>Now here’s the challenge. IT providers’ classical play lays in Level 4. It is a typical “IT-play” for a data company to connect to level 4 transactional systems, extract data (which very often is organised in tabular formats), offload data to large analytical systems and allow to apply advanced analytics, GenAI etc to large datasets without impacting transactional data. Yet, the source of data already is to some extend aggregated and usually does not contain detailed data that would be produced on Level 1, 2 or 3 of the ISO-95 pyramid. Another challenge is, that we usually see “point to point” connections between those levels. Think of a sensor, that is connected to a PLC, and a PLC is connected to a SCADA. If there is any change of the sensor (swapping the sensor with a newer device, that reports in different time or data formacts), what one would have to do to reflect this change in the physical world would be a change of the configuration on the higher levels. This makes data management relatively challenging (and this is only one of 1000s of possible examples) as we talk about n-n connections (and we all know how complex this can become).</p><p><strong>UNS — now what?</strong></p><p>Now this is the moment where “UNS” — Unified Name Space — comes into play. Unified Name sSace basically is a centralized, shopfloor centric data management solution. Rather than going from layer to layer, UNS consolidates data form all kinds of sources across those layers, adds semantic context (which sensor, with which naming conventions, which data points with which data type, which time frames, which measurements, and also relations like in which systems does the sensor sit, in which factory does the system sit etc). It also allows to “abstract” differences coming from different vendors (well, at least from a data point of view). Scalability is another advantage, also looking across factory borders. So basically, UNS offers a “factory data model”.</p><p>UNS conceptually is close to a well organized analytical database setup. The big difference is, that UNS systems usually sit on the shopfloor, and shopfloors are often “DMZs” — demilitarized zones, with extremely strongly governed internet access policies, which makes a central configuration from a cloud infrastructure often a challenge.</p><p>Snowflake can adress those challenges via various partners, that provide UNS, shopfloor based solutions, and provide outbound connectivity usually via MQTT. This allows a very controlled configuration of outgoing traffic from the shopfloor. Our partner Cirrus Link for examples integrates with Snowflake in a way, that the factory data model (that has been built on the shopfloor UNS system) is basically “replicated” in Snowflake tables without the need for additional modelling; tables are created and populated (based on MQTT messages) basically automatically. Every new MQTTS message (that, e.g. represents a status change of a sensor), poplulates the tables. The combination with Snowpipe Streaming allows a cost efficient setup, as no warehouse has to be up and running for ingestion purpuses, but is rather triggered “on demand” as a serverless service (which is immediately fired up and shut down as needed). This addresses a common, yet unnecessary concern of high costs for ingestion.</p><p><strong>What’s next?</strong></p><p>Snowflake is continuously widening its partner ecosystem and ingrestion technologies to offer flexible solutions. Besides Cirrus Link, we also work with HighByte, Crosser, allow Kafka based connectivity to name but a few. I am personally also very keen in exploring, what the recent aquisiton of Datavolo will bring in terms of additional value and flexibility moving forward.</p><p>Next question: what do you do with the data once you have it on your data platform? I will provide updates of built-in Snowflake capabilities, that significantly accelerate time-to-value when working with shopfloor data.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0cc656e70a69" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>