Ozgur Tohumcu writes about applying AI, cloud technology and automation to manufacturing, automotive and supply-chain operations. Tohumcu's posts connect industry trends and customer examples with AWS products and services.
Physical AI robots learn task by task and environment by environment, not from internet data.
Scaling factory AI needs connected manufacturing data, model and compute access, production safeguards.
Volkswagen's platform spans 43 plants; TVS Motor cut line stoppages 3% to 5%.
Check first: Confirm audience demographics, recent post results, video performance, pricing and sponsorship terms.
Main ideas
Industrial AI matters when it moves beyond pilots and into operations · Physical AI has to learn through real tasks and environments · AI and automation should serve the product and the factory’s needs
Ozgur Tohumcu writes about applying AI, cloud technology and automation in manufacturing, automotive and supply chains.
In LinkedIn posts, they connect industry developments and customer examples to practical questions: how to move AI from pilots into production, what data and infrastructure that requires, and where physical AI could change factory operations.
Field notes from Brazil, Japan and industry events add local observations to that technical and business focus.
To assess the depth of their work beyond these recurring themes, ask for examples of a signature book, show or earlier body of work.
Tohumcu explains industrial AI through operational constraints and decisions, not as a stand-alone technology trend.
On physical AI, they note that robots cannot rely on a vast store of internet training data: they must learn task by task and environment by environment, with closed-loop learning allowing one machine’s experience to improve others.
For scaling factory AI, they offer a practical checklist—connected manufacturing data, access to models and compute, and safeguards for production use—and cite Volkswagen’s platform across 43 plants as an example.
Customer-visit posts bring a field perspective: in Brazil, for instance, they describe discussions focused on yield, downtime and production efficiency, and argue that local presence matters alongside global expertise.
Readers get concrete ways to think about industrial AI adoption, while the AWS product and service context calls for distinguishing the operational arguments from the vendor’s commercial framing.
Ozgur Tohumcu’s posts frame industrial AI as an operating challenge, not just a model choice: manufacturers need sound data foundations, production-ready deployment and a way to learn from use in the physical world.
Read more
They make the case through customer and market examples, repeatedly tying the argument to AWS products and partnerships.
Industrial AI matters when it moves beyond pilots and into operations
Tohumcu argues that pilots alone do not deliver AI’s full value.
Read more
Scaling requires connected, AI-ready manufacturing data, access to models and computing capacity, and safeguards suitable for production. The examples focus on operational outcomes, not AI for its own sake.
In a post about a HANNOVER MESSE discussion with Volkswagen’s Florian Lechner, Tohumcu says Volkswagen’s Digital Production Platform has unified 43 plants on one factory cloud. They present data foundations, freedom to innovate and trustworthy deployment as requirements for moving from pilots to scale.
Physical AI has to learn through real tasks and environments
Tohumcu distinguishes AI that works with digital information from AI that controls machines in the real world.
Read more
Physical AI faces different data and operating constraints.
With no ready-made internet-scale dataset of manipulation, systems must be developed task by task and environment by environment.
Tohumcu sees fleet learning as a route to scale: experience from one deployed machine can improve others.
Tohumcu describes closed-loop learning as a way for a robot that solves a new problem to improve other robots in the network. Another post points to factory tasks such as robotic arms learning new assembly work and autonomous systems inspecting or moving materials.
AI and automation should serve the product and the factory’s needs
Tohumcu puts business and production goals ahead of adopting technology for its own sake.
Read more
They also see manufacturers making deliberate choices about which capabilities differentiate them and should stay in-house, and which are better handled with partners.
After meeting automotive and manufacturing leaders in Tokyo and Nagoya, Tohumcu writes that desirable products come first and technology serves the product. The post also describes clearer decisions about which parts of the software-defined vehicle stack to keep in-house and which to develop with partners.
Supply-chain AI should help teams act before disruptions become crises
Tohumcu says supply chains are too large and fast-moving for people to monitor every signal manually.
Read more
In their account, AI agents can scan operational data, find causes and recommend or carry out approved actions, shifting teams from reactive firefighting toward earlier decisions. The examples are tied to a product launch and named customers.
In a post announcing Amazon Connect Decisions, Tohumcu describes agents scanning supply-chain data and learning from planners.
They cite TVS Motor’s reported 3–5% reduction in line stoppages, with 15–20% expected at full rollout, and Wells Vehicle Electronics’ reported 7% inventory reduction in three months.
They say the system works with existing ERP systems.
Local knowledge and sustained presence matter in industrial markets
Tohumcu argues that global expertise alone is not enough to earn trust: companies need people who understand local industries and can connect local conditions to broader experience.
Read more
They describe industrial AI adoption in practical terms such as yield, downtime and product validation, and stress that credibility takes time to build.
After five days in Brazil, including visits with eight enterprise customers and two factories, Tohumcu says showing up matters more than sending a deck. They describe AI conversations focused on production outcomes and credit a local team’s two years of work with building credibility among cautious customers.
Start with a first-hand account of a platform launch, the customer problems behind it and the partners already building on it.
The post announces the launch of the Physical AI Toolchain, an open-source stack on AWS built with NVIDIA's physical AI stack to cover the development lifecycle for machines that perceive and act in real-world settings.
Two themes recur in customer conversations: use cases are already here (robot arms taking on new assembly tasks without reprogramming, autonomous systems inspecting and moving materials, and factories spotting production issues before they happen); and value compounds with each deployment as machines in the field generate operational data that improves models across the fleet.
NEURA Robotics, RLWRLD and Config are named as builders on AWS.
For brands planning a campaign
A robotics, machine-vision or factory-automation vendor can sponsor or co-create a launch-timed post, with the fleet-learning argument offering a ready angle for a case study, demo or webinar.
What is this creator's technical argument, apart from a product announcement?
Read more
This is the clearest short explanation of how physical AI differs from the factory AI that came before it.
The post argues that physical AI relies on different data, architectures and constraints from the AI already used in manufacturing.
Tohumcu notes there is no 'internet of manipulation data' to train on, so capability must be built task by task and environment by environment.
Closed-loop learning is the mechanism that makes it scale: every robot that solves a new problem improves every other robot in the network, changing the economics at volume.
For brands planning a campaign
Useful for technical audiences that need the category defined before a pitch. A data, simulation or robotics vendor could sponsor or respond to this framing, which names the training-data bottleneck buyers ask about.
Why do so many industrial AI projects stall, and who is getting past it?
Read more
Drawing on a stage conversation with a named manufacturer, this post answers both with a short checklist buyers can reuse.
The post opens with the claim that 85% of manufacturers are still stuck in AI pilots, then lists three requirements discussed on stage with Volkswagen's Florian Lechner, head of digital production and logistics: an AI-ready data foundation that unifies manufacturing data on a secure, scalable cloud; access to hundreds of foundation models and the compute to use them; and trustworthy deployment with guardrails, policies and enterprise-grade security built for production. Volkswagen's Digital Production Platform is cited as unifying 43 plants on a single factory cloud, with gains in throughput, quality and sustainability.
For brands planning a campaign
A vendor selling manufacturing data platforms, MLOps or AI governance can build on the three-part checklist rather than use a generic AI message. The named 43-plant example gives readers a concrete benchmark.
What results are being claimed for agentic AI in supply chains, and what evidence backs them?
Read more
This launch post cites named customers and specific numbers rather than general promises.
The post announces Amazon Connect Decisions, an agentic supply-chain planning solution built on more than 30 years of Amazon operational science managing over 400 million SKUs.
It cites two customer results: TVS Motor reports a 3% to 5% reduction in line stoppages, with 15% to 20% expected at full rollout, while Wells Vehicle Electronics cut inventory 7% in three months and drove fill-rate penalties to zero.
Tohumcu describes a shift from reactive firefighting to decisioning AI teammates that learn from planners, detect invisible patterns and execute approved actions around the clock, working with existing ERP systems rather than replacing them.
For brands planning a campaign
For a supply-chain software brand, this offers a template for adoption claims in the field: named customers, staged numbers and reassurance about integration. It also creates a natural counterpoint for a competing or complementary planning tool.
What does this creator notice in face-to-face meetings with manufacturers, and how do they build the argument?
Read more
This is the fullest field-note post in the set.
Tohumcu reports a week in Tokyo and Nagoya with automotive and manufacturing leaders, OEMs, suppliers and partners, drawing five takeaways.
Desirable products come first, illustrated by Toyota's roughly $323B record fiscal-year revenue, a seventh straight year leading global sales, Japanese sales up about 6% and electrified vehicles now over half of what it sells.
Teams are getting sharper about which parts of the software-defined vehicle stack to keep in-house and which to partner on.
The GPU crunch in Japan is real: demand outstrips supply, and the weak yen squeezes start-up infrastructure budgets.
Choosing an AI partner has become a CEO and board decision, not a delegated one.
Physical AI is moving from idea to planned factory-floor scale, with the humanoid market forecast to grow from about $3B in 2025 to more than $15B by 2030.
The post closes with how the local team worked with customers.
For brands planning a campaign
Automotive and industrial suppliers get a market read with quotable numbers, while the team-first close suits employer-brand or regional partnership campaigns. The in-house-versus-partner framing offers a ready hook for a technology partner pitch.
How does local market detail become a global argument?
Read more
This post shows the method through named factories, companies and sectors.
The post covers 1,690 miles across two Brazilian states, eight enterprise customers, two factory tours and a full day of executive visioning.
Brazil is described as one of the highest-potential industrial markets, citing its position as the seventh largest vehicle market and Latin America's biggest manufacturing economy.
Examples span heavy trucks and buses (Marcopolo), aerospace (Embraer), precision electronics (Elgin) and consumer goods (3 Corações).
The three reflections: showing up matters more than sending a deck, because connecting local observation with global patterns builds trust; AI conversations focused on yield, downtime, production efficiency and product validation rather than theory, since companies competing globally have little margin for experimentation; and credibility with cautious buyers took two years, not quarters, to build.
For brands planning a campaign
A brand entering Latin American industrial markets gets talking points and an argument for local presence, while the named companies give sponsors specific sectors to align with. The point that these technologies are 'in production now, not experimenting' supports a proof-led rather than awareness-led campaign.
The post frames the message for a cross-industry conference audience and highlights the numbers presented.
Read more
At VivaTech in Paris, Tohumcu describes conversations across automotive, manufacturing, energy, aerospace, AI, policy and start-ups that converged on digital transformation, sovereignty and sectoral collaboration.
The recurring point: strong AI models are not enough when the data infrastructure underneath is fragmented; the fastest-moving companies treat data as a strategic asset, with AI as the accelerant.
Supply-chain figures include a 10% to 15% cost reduction and a 20% to 30% improvement in forecast accuracy from optimised supply chains, alongside 35 years of Amazon supply-chain expertise.
The post also mentions a panel on agentic AI for industrial supply chains with Valeo's Rémi Deleaune and an executive dinner credited to named colleagues.
For brands planning a campaign
An enterprise data or integration vendor gets a conference-season message with quotable percentages. A sponsored slot would need to add technical depth: the event summary stays at the level of themes and headline figures.
Field reporting from customer visits and industry events5
Scaling industrial AI past the pilot stage4
Physical AI and robots on the factory floor4
Data and cloud foundations as the real bottleneck4
Named customers and partners as proof points4
Teaching through distinctions and prerequisites4
Agentic AI for supply chain planning3
Scaling industrial AI past the pilot stage
A recurring argument across the posts is that most manufacturers are still running limited AI pilots and that value appears only when AI is deployed across operations.
Read more
They cite the figure that 85% of manufacturers are stuck in pilots, describe moving beyond isolated pilots into demand forecasting and quality control, and report that manufacturers in Brazil are already putting AI into production because they compete globally. They frame the requirements as an AI-ready data foundation, access to foundation models and compute, and governance controls suitable for production environments.
Tohumcu typically structures these posts as numbered lists of requirements or takeaways, anchoring each point to a named company or plant.
For brands planning a campaign
A vendor selling data platforms, MLOps, compute or AI governance tooling can evaluate Tohumcu for content explaining what separates a pilot from a production deployment. The piece can name concrete prerequisites and use a customer example rather than rely on general enthusiasm.
Tohumcu treats Physical AI as a different problem from the software AI already used in manufacturing: it runs on different data and architectures, has no shared training corpus comparable to text, and must be built task by task and environment by environment.
Read more
They point to closed-loop learning, where every robot that solves a new task improves the rest of the fleet, and to humanoid robots as a way to work within existing factory layouts instead of rebuilding them. They announced the Physical AI Toolchain built on AWS with NVIDIA and named NEURA Robotics, RLWRLD and Config as companies building on it.
Tohumcu typically opens with a conceptual distinction, then moves to a product launch or partner example that demonstrates it.
For brands planning a campaign
Robotics, automation, sensor and simulation vendors have a natural fit here. Tohumcu explains the category and names partners in the same post, giving a campaign room to place a customer story within that explanation.
Supply chain is one of Tohumcu's most repeated subjects.
Read more
They describe automotive supply chains as too large and fast-moving for any team to monitor end to end, then present agentic AI as a way to continuously scan data, trace root causes and recommend actions before problems escalate.
Around the Amazon Connect Decisions launch, they cited TVS Motor Company seeing a 3-5% reduction in line stoppages, with 15-20% expected at full rollout, and Wells Vehicle Electronics cutting inventory 7% in three months.
They also stress that the tools work with existing ERP systems rather than replacing them, and describe the shift from AI assistants to AI agents.
Product launch posts consistently pair a named customer with a quantified result.
For brands planning a campaign
Supply-chain planning, ERP integration and operations software vendors can evaluate Tohumcu for launches that need a plain explanation of agentic AI alongside credible customer proof. The quantified results they cite give buyers details they can check.
Tohumcu repeatedly argues that the AI model is not the constraint; the data and infrastructure underneath are.
Read more
They call data "the connecting tissue," say great models do not help if the data infrastructure is fragmented, and list an AI-ready data foundation as the first requirement for scaling. They connect this to cloud capacity, including a real GPU crunch in Japan that is squeezing startups whose budgets are consumed by infrastructure costs, and to Amazon's long-term capacity investment in the region.
For brands planning a campaign
Data platform, cloud infrastructure and compute providers can use Tohumcu for content that reframes an AI conversation around its foundations. The argument is concrete enough to engage a technical audience without becoming a product pitch.
Field reporting from customer visits and industry events
Much of the material is first-person reporting from travel and events: five days in Brazil covering 1,690 miles, eight enterprise customers and two factories; a week in Tokyo and Nagoya with automotive and manufacturing leaders; VivaTech in Paris; and Hannover Messe.
Read more
The format is a numbered list of observations, tying specific local detail to a broader industry pattern, followed by credit to teams and colleagues by name. Posts typically open with scene-setting numbers, move through three to five numbered lessons, and end with a forward-looking line.
Named people and teams appear regularly, including customer executives, panelists and colleagues; the posts usually close by thanking the local team.
For brands planning a campaign
Brands seeking event-based or field-report content can evaluate Tohumcu for on-the-ground accounts that specify geography, distance and meeting counts rather than offering abstract commentary from a desk.
Tohumcu's posts name specific organisations and quantify outcomes: Volkswagen's Digital Production Platform unifying 43 plants onto a single factory cloud; TVS Motor and Wells Vehicle Electronics on supply chain results; NEURA Robotics, RLWRLD and Config building on AWS; and a panel with Valeo on agentic AI for industrial supply chain.
Read more
They also use Amazon's own operations as credentials, citing 35 years of supply chain expertise, 400M+ SKUs and one million robots in operation.
Every post reviewed is framed around AWS products, launches or partnerships, and customer examples serve that story. A vendor outside that ecosystem would be working against the grain of the existing material rather than extending it.
For brands planning a campaign
Enterprise software and industrial technology vendors can evaluate Tohumcu for content that needs third-party proof rather than opinion. The posts regularly name companies, so vendors seeking a specific customer example can request one.
Several posts hinge on a clarifying distinction rather than a news hook: Physical AI is not a continuation of existing manufacturing AI; a pilot is not a production deployment; an AI assistant is not an AI agent; showing up is not the same as sending a deck.
Read more
Tohumcu then lists what follows from each distinction. The approach suits readers facing a technical or organisational decision who need the categories separated before considering a recommendation.
For brands planning a campaign
Vendors whose products need a category explained before they can be sold can use Tohumcu for that work. The posts show they are willing to state what a technology is not, giving a campaign a sharper angle than a general capability overview.
Tohumcu writes from inside Amazon Web Services, describing AWS launches, teams and customer results in the first person.
Read more
The other named companies appear as AWS customers or technology partners in their examples. Ask for confirmation of any sponsorship, affiliate or investment relationships and their job title.
Amazon Web Services (AWS)
Employment
Tohumcu's posts speak from inside AWS.
Read more
They refer to "our team" in Japan and Brazil, "our own" colleagues, "our ongoing investment in capacity" and "our learning across the industry," and introduce AWS products such as Amazon Connect Decisions and the Physical AI Toolchain as launches they are part of. Ask for their job title and start date.
Tohumcu's HANNOVER MESSE post names Volkswagen as an AWS customer example: its Digital Production Platform on AWS unifies 43 plants on a single factory cloud.
Read more
The post identifies Volkswagen as an AWS customer, not as a brand they personally work for.
The post names Wells Vehicle Electronics as cutting inventory 7% in three months and driving fill rate penalties to zero with Amazon Connect Decisions.
Read more
It presents the company as an AWS customer result.
Ozgur Tohumcu writes about industrial AI, physical AI and supply chain from the standpoint of someone selling and deploying the technology inside a large cloud vendor.
Read more
Most posts are first-person field reports: a week with automotive and manufacturing leaders in Tokyo and Nagoya, five days in Brazil, VivaTech in Paris and Hannover Messe.
They follow a consistent structure: scene-setting numbers, three to five numbered takeaways, then credit to the teams and colleagues involved.
Follow the work by thread.
For manufacturing AI, start with the argument that most manufacturers are stuck in pilots and that data foundations, access to models and production-grade governance decide who scales.
For robotics, read the posts that distinguish Physical AI from ordinary software AI and describe closed-loop learning across a robot fleet.
For supply chain, the Amazon Connect Decisions posts pair agentic AI with named customers and measured results.
Across the material, Tohumcu names the company and the number, and almost every example ties back to an AWS product, launch or partner.
Writes from inside industrial automation and engineering about the same ground Tohumcu covers: physical AI, autonomous systems on the factory floor, and what happens when AMRs, cobots, AI agents and humanoids operate together.
Read more
Both treat manufacturing operations as the place to watch the technology land.
Joseph teaches from an engineering-practitioner perspective, reaching back to manufacturing history such as Taiichi Ohno and just-in-time, and closes with open questions rather than answers. His conference and speaking posts carry no vendor product or customer-result framing, while Tohumcu reports deployments and launches, usually ending with a named customer and a number.
Covers physical AI, robotics and the AI infrastructure buildout, including which layers of the physical AI stack attract capital and how software demand propagates into power, cooling and manufacturing suppliers.
Read more
The subject overlap with Tohumcu is close, down to Japan's industrial base and factory data as a training asset.
Yan writes as an investor-analyst, framing market structure, funding rounds and theses about China, Japan and the US. Tohumcu writes as an operator inside a vendor, describing customer deployments and product launches rather than where capital should go.
Tracks AI and robotics in automotive and manufacturing, including BMW's Neue Klasse reset as an AI-first redesign of a car company and a Munich robotics funding round.
Read more
That is the same intersection of cars, factories and AI that Tohumcu covers, with a shared European and German industrial focus.
Kim's posts are strategy and market summaries drawn from corporate announcements, written in a third-person analyst voice and dense with hashtags. Tohumcu's are first-person accounts of visits and events, tied to specific AWS offerings and the teams and colleagues involved.
The only measured audience figure available is 13,153 LinkedIn followers on the profile linked from Ozgur Tohumcu's posts, observed 8 October 2026.
Read more
Their posts do not identify followers by country, industry, job function, or seniority.
Their first-person industry reporting covers VivaTech in Paris (June 2026), HANNOVER MESSE with Volkswagen's head of digital production and logistics (April 2026), a five-day Brazil trip covering eight enterprise customers and two factories (July 2026), and a week in Tokyo and Nagoya with automotive OEMs, suppliers, and partners (October 2026).
Recurring subjects include supply-chain AI, agentic planning tools, and physical AI and robotics on the factory floor, with named companies including Volkswagen, Toyota, TVS Motor Company, and Wells Vehicle Electronics.
The subject matter points to readers working in automotive, industrial manufacturing, supply-chain operations, and industrial-AI programs.
For a manufacturing technology provider, LinkedIn is a natural home for a campaign aimed at manufacturing technology buyers.
Before committing to reach, ask for recent post results—impressions, reactions, comments, and follower demographics such as country, industry, and seniority—for comparable topics, depth, and formats.
If the campaign depends on job function or decision-making authority, ask whether Ozgur Tohumcu has audience survey data, since platform follower data cannot verify occupation or seniority.
The posts publish no video performance data, so ask for comparable results before choosing a format or topic.
Treat the follower total as a starting scale check and the posts as an editorial-fit check.
Ozgur Tohumcu’s LinkedIn posts offer industry commentary from inside AWS: event recaps, customer and factory visits, and product launches such as Amazon Connect Decisions and the Physical AI Toolchain.
Read more
Volkswagen, TVS Motor, Wells Vehicle Electronics, NEURA Robotics, Toyota, Marcopolo and Embraer appear as AWS customers, plant hosts or panel partners—not as brands paying them for sponsored content.
The posts disclose no sponsorships, rates, deliverables or booking terms, and include no media kit, rate card, paid-placement audience breakdown or partnership inquiry page.
Their LinkedIn account is the only public channel attached to the profile, and it is a general professional profile, not a published partnership route.
For campaign planning, treat them as an enterprise industry voice rather than a creator with established placement formats.
Ask directly about availability, scope and terms for a paid arrangement; the posts do not say whether they accept sponsorships.
Campaign need: a manufacturing data, cloud or MLOps vendor wants a launch campaign that reaches automotive and industrial manufacturing decision-makers, with content that explains why AI projects stall at the pilot stage and what a production deployment requires.
Read more
This is the strongest match in the posts.
Ozgur Tohumcu's posts make exactly that argument and carry proof a technical buyer can check: Volkswagen's Digital Production Platform unifying 43 plants on a single factory cloud, TVS Motor Company reporting a 3% to 5% reduction in line stoppages with 15% to 20% expected at full rollout, and Wells Vehicle Electronics cutting inventory 7% in three months.
They also spell out the prerequisites: an AI-ready data foundation, access to foundation models and compute, and governance controls suitable for production environments.
For a vendor selling any of those layers, the message and audience line up.
The reach case is LinkedIn only: LinkedIn lists 13,153 followers as of 8 October 2026.
Ask for an admin export covering comparable topics before committing budget; the posts publish no follower demographics or video performance data.
Nearly every post is framed around AWS products, launches or partnerships, making this the cleanest fit for a vendor working in that ecosystem or in a non-competing layer.
A rival cloud provider would be asking them to argue against the grain of their posts.
When to choose a specialist
Campaign need: a consumer snack brand wants a paid short-form video placement promoting a new product to a broad consumer audience, with a fixed fee, a guaranteed engagement target and a two-week booking window.
Read more
The posts do not support this campaign.
All nine are enterprise industrial commentary — automotive supply chains, factory AI, robotics and cloud infrastructure — for people who make technology or operations decisions, not shoppers.
There is no video, transcript, visual asset or performance data to assess delivery.
The audience is unmeasured beyond the follower count, with no country, industry, seniority or age breakdown.
The posts also provide no rate card, sponsorship disclosure, deliverable list or booking terms, and the profile is a general professional profile rather than a partnership route.
Before considering a booking, ask for audience data, video performance, pricing and terms.
The subject matter is a poor fit; a consumer brand would be better served by a creator with demonstrated short-form video reach.
A worked campaign example
Campaign angle: a robotics or machine-vision vendor launching an automated inspection product for automotive plants.
Structure it as a three-post sequence that mirrors the format these posts already use.
Post one makes a distinction, not a pitch: why automated inspection running on real factory data is a different problem from the vision systems already deployed on production lines.
Tohumcu's post on physical AI does this well, arguing that there is no shared training corpus equivalent to text and that capability has to be built task by task and environment by environment.
Name the constraint openly; it gives the next post somewhere to go.
Post two is a field report from a customer plant, using the repeatable shape of these posts: scene-setting numbers first (miles travelled, sites visited, people met), then three to five numbered takeaways, then a forward-looking line. Each takeaway should attach to a named customer and a measured result rather than a general claim.
Post three is a short checklist of what a plant needs in place before inspection AI moves from pilot to production, with each prerequisite stated in one sentence a reader can reuse in an internal note.
The campaign should ask for a named customer, at least one number, and one explicit statement of what the technology is not. Since no collaboration format is publicly described, agree the commercial terms — scope, deliverables, rights and disclosure — directly before scheduling, and disclose the arrangement on the posts once it exists.
Choose Joshua Joseph when the campaign needs plant-floor practice rather than cloud architecture, and the audience is industrial engineering, automation or robotics-integration teams.
Read more
Their posts cover speaking at Automate Show, Robotics Summit & Expo, Modex and IISE Annual.
They frame the current question as what a factory looks like when AMRs, cobots, AI agents and humanoids run together, while people still make the decisions those systems cannot own.
One post grounds the same argument in an origin story: Taiichi Ohno at a Piggly Wiggly supermarket in 1950, just-in-time, and asking where the waste is.
For vendors explaining how work is divided between people and machines on a real factory floor, that framing is closer to the audience's daily language than a platform argument.
Ozgur Tohumcu's posts do not name a customer, report a measured result or provide product or platform proof.
They argue qualitatively that industrial AI can move from pilot to production, citing Volkswagen's 43-plant factory cloud, TVS Motor's line-stoppage figures and a documented inventory cut.
Vendors seeking pilot-to-production metrics or a data-foundation story should ask for more evidence.
Choose Joe Yan when the campaign targets investors, corporate strategy or supply-chain finance audiences, or when the product sits inside AI's physical build-out: power, cooling, data-centre capacity or industrial components.
Read more
Their posts trace capital moving beyond robot hardware into a modular stack of components, embodiment, motion and control, foundation models and data, and make a regional case for where each layer gets built.
The second post follows AI demand into established industrial suppliers: Generac's data-centre backlog is reported at roughly $1.6 billion, and Modine's long-term cooling capacity agreement is valued at more than $4 billion, with $165 million paid upfront.
This answers the same scaling question through money flows rather than deployments.
This is investment analysis and market thesis, not field reporting.
There are no named customer deployments or measured plant outcomes.
Vendors seeking a factory-grounded pilot-to-production narrative should ask for deployment evidence.
The regional frame—how China, the US and Japan divide the physical AI stack—can also steer a campaign toward macro commentary rather than a specific product discussion.
Choose Hyunjeong Kim when the audience is European OEM strategy, procurement or Tier-1 supplier teams, and the message concerns how a manufacturer reorganises around AI.
Read more
One post examines BMW's Neue Klasse reset in detail: core technology kept globally scalable while development, sourcing and production move closer to regional customers; at least 95% local development and production planned for China; AI applied across engineering, knowledge management, procurement and pricing; a stated plan to cut around 8,000 positions by the end of 2027; and a supplier model that keeps differentiating technology in-house while co-developing standardised parts.
The second post covers the Munich robotics ecosystem, including RobCo's $100 million Series C, its robotics-as-a-service rental model and IFR's projection of more than 700,000 annual robot installations by 2028.
For vendors selling into German or European automotive, this is direct adjacency.
The detailed evidence centres on one OEM and the German and European market, giving vendors seeking proof across many customers or regions a narrower base than Tohumcu's multi-country reports. The material reports corporate strategy, not measured production results; vendors seeking deployment numbers should ask for them.
This profile covers nine LinkedIn posts by Ozgur Tohumcu: product announcements, arguments about physical AI and robotics, manufacturing and supply-chain examples, and first-person reports from customer visits and events including VivaTech, HANNOVER MESSE, Tokyo and Nagoya, Brazil, and US and European plant visits.
Six posts by three other creators provide similar-creator comparisons.
Tohumcu's posts carry no publication dates; the snapshot is dated 8 October 2026, and the comparator posts run from January to October 2026.
Several gaps matter to readers.
These posts are a selection rather than a full record, so earlier work and any signature book, show or career change are not covered.
The posts here include no comments, articles, other channels, video or transcripts, so no format beyond text can be judged on delivery.
Audience is unmeasured beyond a follower count: no country, industry, seniority or engagement data.
No rate card, media kit, sponsorship disclosure or booking terms appear in the posts here, so commercial arrangements are not documented, not ruled out.
One commercial relationship is documented — Tohumcu writes from inside AWS, describing its launches, teams and customer results — while the other companies named are AWS customers or technology partners in the examples.