Manufacturing Hub

Digital twins and machine learning are redefining batch optimization in manufacturing. Learn how centerlining models can catch quality issues in real time before they become irreversible.

Concepts like digital twins, golden batch profiles, and statistical process control have long promised more than they delivered. Virag Vora of Twin Thread argues that layering machine learning on top of these ideas is what finally brings them to life. In this context, a digital twin is entirely data centric: a real time and historical representation of a process that serves as the foundation for AI models.

The core use case is batch centerlining. The model compares current conditions against historically successful profiles, segmented by raw material source, product type, and seasonality. An orange juice manufacturer uses Twin Thread to determine whether incoming fruit should be sold fresh or routed to concentrate based on seasonal sugar content. The model identifies contributing variables in real time and alerts operators before a batch drifts beyond recovery.

Twin Thread tackles the "not enough data" objection head on. With over 60 connectors, the platform works with the fragmented data reality of most manufacturing sites. Even low frequency data can train a useful model that quantifies what higher resolution instrumentation would unlock.

Virag draws a clear line between ML and LLMs for process control. ML models trained on historical data produce deterministic outputs trusted for real time guidance on machine settings. LLMs excel at document retrieval and natural language interaction but are not suited for recommending set points on a live line. Twin Thread layers both: ML handles optimization, while Twin Thread Advisor lets users interrogate data and configure models through conversation.

The standout proof point is Hills Pet Nutrition. After three years on Twin Thread, their models automatically feed recommendations into live production. That closed loop followed a deliberate path from human validation to A/B trials to automated execution with operator opt out.

About Virag Vora
Virag Vora is a solutions professional at Twin Thread, a platform that combines data centric digital twins with machine learning to optimize manufacturing processes. With a background in chemical engineering, Virag began his career deploying MES and DCS systems in biotech and pharma before joining Tulip and then Twin Thread. He helps manufacturers connect their existing data infrastructure to AI powered optimization across batch, continuous, and hybrid processes.

Timestamps
0:00 Introduction
1:20 Virag's background in chemical engineering and industrial software
6:30 Moving up the ISA 95 stack from DCS to MES and applications
9:00 How AI reinvents digital twin, golden batch, and SPC concepts
12:20 What a data centric digital twin actually looks like
21:40 Where digital twins deliver the most value in manufacturing
27:00 Seasonality, segmentation, and model training strategies
36:00 Data prerequisites for deploying industrial AI
41:40 Flavors of AI in manufacturing: ML, LLMs, and agentic workflows
50:40 Closed loop AI control at Hills Pet Nutrition
53:10 Personal project: Family Graph using knowledge graphs
56:20 Prediction: operators as human digital twins

References
Twin Thread: https://twinthread.com

This episode is sponsored by
MaintainX is an AI powered maintenance and operations platform that helps technicians get the answers they need instantly so they can focus on getting assets back online. Learn more about how MaintainX supports frontline manufacturing teams.
https://maintainx.com

About Your Hosts
Vladimir Romanov is a co-host of The Manufacturing Hub Podcast and the founder of Joltek, an independent manufacturing and industrial automation consulting firm specializing in modernization strategy, digital transformation, and workforce development. Joltek works with manufacturers and investors to de-risk modernization and build the internal capability to sustain results.
Connect with Vlad: https://www.linkedin.com/in/vladimirromanov/

Want to go deeper? Vlad and the team at Joltek have covered related topics here:
Edge Computing, AI, and the Value of Manufacturing Data: https://www.joltek.com/blog/edge-computing-ai-value-manufacturing-data
Digital Transformation in Manufacturing: https://www.joltek.com/blog/digital-transformation-in-manufacturing

Dave Griffith is a co-host of The Manufacturing Hub Podcast and founder of Capelin Solutions, an industrial automation firm helping manufacturers adopt smart manufacturing technology. He brings 15 years of experience in industrial automation and digital transformation.
Connect with Dave: https://www.linkedin.com/in/davegriffith23/

Subscribe to Manufacturing Hub: https://www.manufacturinghub.live
LinkedIn: https://www.linkedin.com/company/manufacturing-hub-network
YouTube: https://www.youtube.com/@ManufacturingHub

Physical AI is arriving on factory floors ahead of schedule, and Vention is already deploying it on applications four automation integrators failed to crack.
François Giguère, CTO of Vention, draws a precise line between agentic AI and physical AI. Agentic systems process data and return data. Physical AI controls motion and actuation that produce real world consequences on a factory floor where a hundred percent uptime is the only acceptable standard. Giguère has spent a decade helping build Vention, a platform that lets manufacturers design robotic cells in 3D, program them through natural language, simulate them in a browser, and receive the physical machine shipped in modular components like an industrial kit. With a team of 95 engineers and three years as CTO, he brings a grounded perspective on where AI delivers real value in industrial automation and where it still falls short.
The design, automate, simulate workflow at Vention represents one of the most complete implementations of AI-powered machine engineering currently in production. In the design phase, customers build systems from a modular component library. In the automate phase, an AI agent converts natural language prompts into Python control code for the entire cell including robot arms, conveyors, vision systems, and grippers. The program is validated in simulation before a single component ships. This is made possible by Vention's motion streaming architecture: instead of treating the robot as the master controller the way KUKA KRL does, Vention brings all motion planning, inverse kinematics, forward kinematics, blending, and trajectory optimization into its own software stack. The robot becomes a passive component consuming a motion stream, and the entire machine becomes programmable from a single unified codebase that AI tools excel at generating. Giguère notes that Vention's choice to use Python as the programming language for automation control gives their AI tools a measurable edge over environments built on structured text or ladder logic.
Vention's two physical AI products are GRIP (Generalized Robotics Intelligence Pipeline) and Rapid AI Operator, a modular bin picking application built on top of GRIP. The technology relies on transformer-based foundation models.
About François Giguère
François Giguère is the CTO of Vention, an industrial automation platform where manufacturers design, program, simulate, and deploy robotic systems entirely online. Employee number four at the company, he has contributed to Vention's growth for over 10 years and leads a team of 95 engineers. He holds a background in electrical engineering and real-time embedded software development.
Learn more: https://vention.io
Timestamps
0:00 Introduction and welcome
1:00 François Giguère's background and Vention overview
2:20 How AI spans Vention's internal tools and customer products
4:00 Why embedded and robotics code is harder for AI to generate
7:00 Design, automate, simulate: Vention's three-stage AI workflow
13:50 Motion streaming: one unified controller for all robot brands
18:20 Defining physical AI versus agentic AI
20:10 GRIP pipeline and Rapid AI Operator
22:40 Case study: MacAlpine Plumbing bin picking with foundation models
39:40 Nvidia GTC impressions: agentic AI eclipsing physical AI
46:20 Edge versus cloud: why real-time inference stays on-prem
56:10 Predictions: physical AI roadmap and the VLA timeline
This episode is sponsored by:
MaintainX helps maintenance and operations teams work smarter by putting critical information directly in the hands of technicians. According to MaintainX, technicians spend up to 40 percent of their time searching for answers and responding to radio calls rather than fixing assets.
https://www.maintainx.com
About Your Hosts
Vladimir Romanov is a co-host of The Manufacturing Hub Podcast and the founder of Joltek, an independent manufacturing and industrial automation consulting firm specializing in modernization strategy, digital transformation, and workforce development.
Connect with Vlad: https://www.linkedin.com/in/vladromanov/
Want to go deeper? Vlad and the team at Joltek have covered related topics here:
Industrial Robotics: https://www.joltek.com/blog/industrial-robotics
Edge Computing and AI Value in Manufacturing Data: https://www.joltek.com/blog/edge-computing-ai-value-manufacturing-data
Dave Griffith is a co-host of The Manufacturing Hub Podcast and founder of Capelin Solutions, an industrial automation firm helping manufacturers adopt smart manufacturing technology. He brings 15 years of experience in industrial automation and digital transformation.
Connect with Dave: https://www.linkedin.com/in/davegriffith23/
Subscribe to Manufacturing Hub: https://www.manufacturinghub.live
LinkedIn: https://www.linkedin.com/company/manufacturing-hub-network
YouTube: https://www.youtube.com/@ManufacturingHub

AI in manufacturing is no longer a strategy reserved for the boardroom. It is a tool for the technician on the plant floor, and the results are already showing up in real operations worldwide.
Most digital transformation strategies in manufacturing are built for desk workers on the carpeted side of the building, not the operators and technicians keeping production running on the concrete floor. AI platforms have historically been designed for white collar knowledge workers with time to navigate complex systems, leaving the frontline worker as an afterthought. Nick Haase recognized this gap when building MaintainX in 2018, and it became the foundational design principle behind everything the company built. The result is a platform now serving nearly 14,000 customers across manufacturing, food and beverage, facilities management, and any industry that depends on physical assets staying operational.
The core thesis Nick brings to this conversation is that the person with no purchasing authority and no budget is the single most important factor in whether a digital transformation project succeeds or fails. That person is the frontline technician. Building for that user first required a mobile experience so intuitive that no training was needed, one that met workers in the flow of existing work rather than pulling them out of it. If your team needs a 300 page manual to use the platform, the adoption battle is already lost.
The skilled labor shortage in manufacturing is not a forecast. The United States is projected to have more than 3 million manufacturing jobs unfilled by 2030, driven largely by retirement of experienced workers who have spent decades building institutional knowledge. That knowledge cannot be transferred through a job posting. MaintainX attacks this through AI powered voice note capture at work order closeout. Technicians leave a verbal description of what they found and fixed. The platform transcribes it across any language or accent, standardizes it, and builds a living knowledge base that outlasts the retirements of the people who created it. For organizations with similar equipment across dozens of sites, that knowledge becomes portable across locations and years.
About Nick Haase
Nick Haase is a co-founder of MaintainX, a frontline work execution platform for maintenance, reliability, SOPs, safety, and compliance serving nearly 14,000 customers across manufacturing and other asset-intensive industries. Nick is also the host of The Wrench Factor podcast.
Connect with Nick: https://www.linkedin.com/in/nickhaase/
Timestamps
0:00 Introduction
1:30 Nick Haase and MaintainX Background
7:20 Where AI Fits for Frontline Workers
10:00 What Data Foundations Are Needed for AI
13:30 Why Frontline Adoption Determines Digital Transformation Success
16:40 The Skilled Labor Shortage and Retirement Wave
18:30 Voice Notes and AI Powered Knowledge Capture
25:30 Overcoming Change Management and AI Skepticism
34:50 Guardrails and Safe AI for Industrial Environments
45:10 Embedding AI in the Flow of Work
48:30 AI Agents for Parts Forecasting and Automation
55:50 Predict the Future: Maintenance as a Growth Center
References
MaintainX: https://www.maintainx.com
The Wrench Factor Podcast: https://podcasts.apple.com/us/podcast/the-wrench-factor/id1809000028
Origins of Efficiency by Brian Potter: https://www.amazon.com/dp/B0FJG6ZKKJ
Inductive Automation Ignition: https://inductiveautomation.com
This episode is sponsored by MaintainX
Technicians spend up to 40 percent of their time looking for answers rather than fixing equipment. MaintainX puts AI powered knowledge tools directly in the flow of work so frontline teams get the right information in seconds.
https://www.maintainx.com
About Your Hosts
Vladimir Romanov is a co-host of The Manufacturing Hub Podcast and the founder of Joltek, an independent manufacturing and industrial automation consulting firm specializing in modernization strategy, digital transformation, and workforce development. Joltek works with manufacturers and investors to de-risk modernization and build the internal capability to sustain results.
Connect with Vlad: https://www.linkedin.com/in/vladimirromanov/
Joltek: https://www.joltek.com/blog/digital-transformation-in-manufacturing
Joltek: https://www.joltek.com/blog/root-causes-downtime-industrial-automation
Dave Griffith is a co-host of The Manufacturing Hub Podcast and founder of Capelin Solutions, an industrial automation firm helping manufacturers adopt smart manufacturing technology. He brings 15 years of experience in industrial automation and digital transformation.
Connect with Dave: https://www.linkedin.com/in/davegriffith23/
Subscribe to Manufacturing Hub: https://www.manufacturinghub.live
LinkedIn: https://www.linkedin.com/company/manufacturing-hub-network
YouTube: https://www.youtube.com/@ManufacturingHub

Industrial AI is getting a lot of attention in manufacturing right now, but one of the biggest questions is still the most practical one. How do you turn plant data, process knowledge, and operational constraints into something that actually creates value? In this episode of Manufacturing Hub, Vlad Romanov and Dave Griffith sit down with Konstantin Paradizov of Eukodyne for a detailed conversation on what industrial AI looks like when it is applied by people who understand manufacturing, MES, process improvement, data architecture, and the realities of the plant floor.
What makes this discussion especially valuable is that it does not stay at the surface level. Konstantin shares how his background moved from pharma into food and beverage, how Lean Six Sigma and process thinking shaped his approach, and why many of the best opportunities in manufacturing still begin with understanding the actual workflow before talking about software. The conversation explores a theme that comes up again and again in industrial transformation: the biggest gains often do not come from adding more technology first. They come from understanding the problem clearly, identifying what information matters, validating assumptions with the people doing the work, and then using the right mix of tools to move faster.
A major part of this episode focuses on the real use of AI in consulting and discovery. Konstantin explains how his team uses secure transcription workflows, on premises AI infrastructure, cloud models, masking of sensitive information, iterative validation, and ROI driven reporting to create high value outputs in a fraction of the time that would have been required even a year or two ago. This is an important point for manufacturers, system integrators, software teams, and plant leaders. AI is not just something that sits in front of an operator as a chatbot. It can be used behind the scenes to accelerate analysis, strengthen recommendations, shorten discovery, improve documentation, and reduce the cost of getting to a better answer.
The technical section of this episode is especially strong for anyone working in industrial automation, OT data systems, or applied AI. The discussion covers on premises compute, Nvidia based edge hardware, Linux environments, Docker containers, RAG workflows, vector databases, knowledge graphs, MQTT pipelines, HiveMQ, Mosquitto, n8n, Claude Code, Cursor, Gemini, OpenRouter, and the tradeoffs between frontier models in the cloud and smaller or open models deployed closer to the process. One of the clearest takeaways is that manufacturers should not start with the biggest model or the most exciting headline. They should start with the problem, the constraints, the data path, and the economics of the solution.
Vlad also pushes on an issue that matters to almost every manufacturer trying to prepare for AI. If you collect massive amounts of plant data into historians, cloud platforms, and enterprise systems, is that enough to create value later? Konstantin’s answer is thoughtful and realistic. More data alone does not automatically lead to better outcomes. You still need filtering, context, prioritization, architecture, and a disciplined way to separate signal from noise.
Learn more about Joltek here:
https://www.joltek.com/services
https://www.joltek.com/services/service-details-it-ot-architecture-integration

Connect with our guest:
Konstantin Paradizov
https://www.linkedin.com/in/konstantin-paradizov/
Learn more about Eukodyne:
https://eukodyne.com/

Follow Manufacturing Hub for more conversations on industrial AI, digital transformation, OT architecture, SCADA, MES, industrial data strategy, systems integration, and the future of manufacturing technology.
Timestamps
00:00 Welcome and introduction to industrial AI applications
01:50 Konstantin’s background from pharma to manufacturing
05:30 Why food and beverage offered major process improvement opportunities
08:10 How to identify the right manufacturing opportunities to pursue
13:10 Using AI to accelerate discovery, documentation, and customer value
21:20 The on premises AI hardware stack and model selection strategy
30:10 Why iterative validation still matters more than a first AI answer
39:00 Claude Code, developer workflows, and practical AI tool stacks
48:20 On premises versus cloud AI and how to think about the tradeoff
54:10 Small models, low cost hardware, and edge deployment realities
01:05:00 Plant data, historians, filtering, and separating signal from noise
01:14:50 Predictions for industrial AI, career advice, and final recommendations
References and resources mentioned in the episode
MaintainX
https://www.maintainx.com/

Solve for Happy
https://www.mogawdat.com/books
George Orwell 1984
https://www.penguinrandomhouse.com/books/326569/1984-by-george-orwell/
George Orwell Animal Farm
https://www.penguinrandomhouse.com/books/561805/animal-farm-by-george-orwell/

Manufacturing Hub is back with Episode 252, where co hosts Vlad Romanov and Dave Griffith break down what an AI survival guide should actually look like for manufacturing and industrial automation professionals. This is not a hype conversation about replacing people with magic software. It is a grounded discussion about what AI tools can do today, where they fail, why context and data quality matter so much, and how industrial teams should think about experimentation without losing sight of real operating constraints.
In this episode, Vlad and Dave unpack the evolution many engineers and technical leaders have already felt in real time, from early prompt engineering, to agent based workflows, to MCP servers, skills, context management, and the growing cost of tokens and infrastructure. The conversation moves beyond generic AI commentary and into the reality of plant floor environments, where success depends on process knowledge, data architecture, OT constraints, cybersecurity, governance, and clear business value. One of the strongest themes throughout the episode is that manufacturers cannot skip the hard work of structuring data, understanding workflows, and defining use cases simply because AI tools are moving quickly.
Vlad brings a very practical industrial lens to the discussion. Drawing on years of hands on experience across controls, manufacturing systems, plant modernization, and digital transformation, he explains why industrial AI has to start with operational context. A maintenance team, an engineering team, and a quality team do not need the same data, do not ask the same questions, and should not be handed the same AI workflows. That distinction matters. This conversation also highlights why the best industrial AI implementations will likely come from teams that combine domain expertise with strong technical execution, rather than generic AI shops trying to force a solution into environments they do not fully understand.
Dave adds an important systems and adoption perspective, especially around cost, scaling, management expectations, and the danger of trying to prompt your way past foundational architecture work. Together, Vlad and Dave explore why manufacturers are interested in AI, why many are afraid of being left behind, and why so many projects still stall once they hit the realities of obsolete equipment, weak data models, fragmented systems, and unclear ownership of information. They also discuss deterministic logic versus LLM behavior, reporting workflows, industrial dashboards, PLC code generation concerns, and the practical question every manufacturer should ask before investing: what problem are we solving, for whom, and what is the measurable return?
For those new to Vlad, he is an electrical engineer and manufacturing leader with deep experience across industrial automation, controls, data systems, OT architecture, modernization strategy, and plant operations. Through Joltek, Vlad works with manufacturers on digital transformation, IT OT architecture and integration, modernization planning, operational improvement, and technical workforce enablement. Learn more here:
Joltek: https://www.joltek.com IT OT Architecture and Integration: https://www.joltek.com/services/service-details-it-ot-architecture-integration
If you are a plant leader, controls engineer, systems integrator, OT architect, SCADA or MES practitioner, or simply someone trying to separate useful AI workflows from noise, this episode will give you a much more realistic framework for thinking about industrial AI adoption.
Timestamps
00:00 Welcome back and why this episode matters
01:00 Setting up the industrial AI theme for the coming weeks
03:10 From prompt engineering to structured AI workflows
05:30 AI agents, parallel workflows, tokens, and context windows
09:00 MCP tools, Playwright, and what new integrations unlock
16:20 How Vlad researches AI and where useful information actually lives
22:00 Real manufacturing problems versus AI in search of a problem
29:40 Why industrial data architecture is harder than most people think
37:00 OT expertise, workforce enablement, and who should build solutions
45:40 Practical advice for manufacturers starting the AI journey
50:30 Data governance, hallucinations, infrastructure, and cybersecurity
57:20 What looks promising today in reporting, dashboards, and industrial applications