Former Common Robots president Kim Povlsen was named CEO of quantum expertise chief Bluefors. | Credit score: Common Robots
Kim Povlsen, the previous president of Common Robots, yesterday was named the CEO of Bluefors. Primarily based in Helsinki, Finland, Bluefors is a number one developer of cryogenic measurement methods, cryocoolers and different cryogenic product strains for quantum expertise, basic physics analysis and different industries comparable to life sciences and clear vitality.
Teradyne Robotics introduced the management change final week, saying Povlsen left Common Robots (UR) to pursue and exterior profession alternative. He joined UR, the world’s main developer of collaborative robotic arms, in March 2021. Jean-Pierre Hathout, president of fellow Teradyne Robotics subsidiary Cellular Industrial Robots (MiR) changed Povlsen as president of UR.
Earlier than UR, Povlsen has held a number of management positions at Schneider Electrical, a worldwide high-tech firm, the place he led enterprise transformation and expertise methods in Europe, the US, and Asia.
Teradyne not too long ago introduced the monetary outcomes for its first quarter of 2025. Its robotics income was $69 million in Q1 2025, which is down from $98 million in This autumn 2024. This additionally marked a 21% drop in gross sales in Q1 12 months over 12 months. Teradyne Robotics laid off 10% of its world workforce in January 2025.
In his new position, Povlsen will lead Bluefors world crew of 700+ staff to ship cooling methods for quantum expertise.
“I’m excited to hitch Bluefors on its mission to speed up the quantum expertise breakthrough,” stated Kim Povlsen. “Bluefors has a confirmed monitor file of efficiently co-creating merchandise with world expertise firms. Because the market chief of their discipline of cooling options for Quantum Know-how and Low Temperature Physics Analysis, they’ve achieved an undisputed position as the corporate who delivers transformational expertise — making real-world, large-scale quantum computing accessible at this time. I see large alternatives to additional strengthen Bluefors because the benchmark for the industries that they serve.”
Jonas Geust, former CEO of Bluefors, stepped down on the finish of 2024 for private causes. Bluefors founder Rob Blaauwgeers served as interim CEO.
“We significantly worth Kim’s capacity to mix strategic considering, buyer orientation and drive for operational excellence,” stated chairperson of the Bluefors Board Kimmo Alkio. “He’s the catalyst of efficient choice making, a frontrunner who motivates change by speaking overtly. Buyer-centricity is in his DNA and meets Bluefors’ philosophy completely: designing and producing merchandise that prospects want and wish – along with the shoppers themselves – is likely one of the cornerstones of our working mannequin.”
UR this week launched the UR15, its quickest cobot ever. The UR15 has a most TCP pace of 5 m/s to scale back cycle instances, enhance productiveness, and cut back prices throughout functions and industries.
UR additionally introduced this week that it surpassed 100,000 cobots offered over the lifetime of the corporate.
Attendees will achieve the most recent insights from specialists in robotics and AI on cutting-edge analysis, business traits, and rising applied sciences. They’ll additionally find out about investments and matters associated to working a robotics enterprise.
In the event you’re a founding father of a robotics startup, Pitchfire is a good alternative to showcase your concepts and propel your organization to new heights. The competitors takes place from 4-4:45 p.m. PT on Oct. 15 on the Santa Clara Conference Middle.
Startups will ship five-minute pitches to a panel of judges describing their options, enterprise fashions, worth propositions, and extra. The judges will decide a winner and two runners-up primarily based on the startups they imagine to be greatest primed for industrial success. The RoboBusiness Pitchfire first-prize winner takes dwelling $5,000.
Taking part startups obtain a free sales space on the expo flooring, can schedule conferences with buyers in attendance, community with business leaders, and have an open invitation to look on The Robotic Report Podcast.
Erin Linebarger, co-founder and CEO of Robotics 88, through the 2024 RoboBusiness Pitchfire Startup Competitors.
Tips on how to apply to 2025 RoboBusiness Pitchfire
Startups should meet the next standards to be eligible for this 12 months’s competitors:
At present creating or commercializing robots or associated applied sciences
Lower than 5 years previous
Fewer than 30 staff
Unfunded or has obtained Seed funding
The deadline for submission is July 16, 2025. To submit your organization as a possible RoboBusiness Pitchfire participant, fill out the entry type beneath or right here. In the event you’d wish to be a choose for the occasion, please contact Steve Crowe: scrowe AT wtwhmedia.com.
Previous Pitchfire winners
Robotics 88 gained the 2024 Pitchfire competitors for its use of drones in wildfire administration. Erin Linebarger, co-founder and CEO of Robotics 88, joined Episode 182 of The Robotic Report Podcast. She mentioned the genesis of the Boston-based firm, the significance of knowledge in fireplace administration, the expertise behind its drones, and the way forward for wildfire mitigation efforts globally. The judges had been impressed not solely with the element of her marketing strategy, but in addition with the corporate’s potential with robotics for good.
Glidance gained 2023 Pitchfire competitors. The corporate’s flagship product, Glide, gives autonomous mobility help for the visually impaired, incorporating superior guiding applied sciences and a singular mechanical design to foster independence. CEO Amos Miller, himself blind, brings a firsthand perspective to Glidance’s mission.
Within the U.S., round 1 million adults are blind. But, solely 2% to eight% use a white cane for navigation. Most as a substitute depend on information canine or sighted companions, in keeping with the Perkins Faculty for the Blind. This reliance limits independence and mobility, a problem Glidance is addressing by robotics.
Tatum Robotics was the 2022 Pitchfire winner for its robotic hand-signing answer for deaf-blind people. By its participation in Pitchfire, the corporate founder Samantha Johnson was invited to inform her story on The Robotic Report Podcast. She shared her firm’s journey to serving to enhance the lives of an typically remoted person base that now has a way more related future within the palms of their arms.
Melting sea ice modifications not solely how a lot gentle enters the ocean, but additionally its colour, disrupting marine photosynthesis and altering Arctic ecosystems in refined however profound methods.
As international warming causes sea ice within the polar areas to soften, it’s not simply the quantity of daylight coming into the ocean that modifications. The colour of the underwater gentle shifts, too, and this has main penalties for all times beneath the floor.
In keeping with new analysis revealed in Nature Communications, these modifications may considerably impression tiny however important organisms like ice algae and phytoplankton. The examine was led by marine biologists Monika Soja-Woźniak and Jef Huisman from the College of Amsterdam’s Institute for Biodiversity and Ecosystem Dynamics.
The worldwide group of scientists, which included bodily chemist Sander Woutersen and collaborators from the Netherlands and Denmark, explored how melting sea ice transforms the underwater gentle setting. Mild behaves very in a different way in sea ice in comparison with open water.
Sea ice displays and scatters most daylight, letting solely a small quantity by means of, however that small quantity contains almost all seen wavelengths. Open seawater, alternatively, absorbs reds and greens, whereas permitting blue gentle to journey deeper. This is the reason the ocean seems blue to our eyes.
Molecular vibrations of water
One other key distinction between ice and liquid water lies within the position of molecular vibrations. In liquid water, H₂O molecules are free to maneuver and vibrate, which ends up in the formation of distinct absorption bands at particular wavelengths. These bands selectively take away parts of the sunshine spectrum, creating gaps within the gentle out there for photosynthesis.
Earlier analysis by Maayke Stomp and Prof. Huisman demonstrated that these molecular absorption options create ‘spectral niches’—distinct units of wavelengths out there for photosynthetic organisms. Phytoplankton and cyanobacteria have advanced a variety of pigments tuned to the totally different spectral niches, shaping their international distribution throughout oceans, coastal waters, and lakes.
Measurements below sea ice by one of many Danish researchers on Greenland. Credit score: Lars Chresten Lund-Hansen
In ice, nonetheless, water molecules are locked right into a inflexible crystal lattice. This mounted construction suppresses their potential for molecular vibrations and thereby alters their absorption options. As a consequence, ice lacks the absorption bands of liquid water, and therefore a broader spectrum of sunshine is preserved below sea ice. This elementary distinction performs a key position within the spectral shift that happens as sea ice melts.
Ecological implications
As sea ice disappears and offers solution to open water, the underwater gentle setting shifts from a broad spectrum of colours to a narrower, blue-dominated spectrum. This spectral change is essential for photosynthesis.
“The photosynthetic pigments of algae residing below sea ice are tailored to make optimum use of the big selection of colours current within the little quantity of sunshine passing by means of ice and snow,” says lead writer Monika Soja-Woźniak. “When the ice melts, these organisms abruptly discover themselves in a blue-dominated setting, which gives a lesser match for his or her pigments.”
Utilizing optical fashions and spectral measurements, the researchers confirmed that this shift in gentle colour not solely alters photosynthetic efficiency, however can also result in modifications in species composition. Algal species specialised in blue gentle could achieve a robust aggressive benefit compared to ice algae.
In keeping with Prof. Huisman, these modifications can have cascading ecological results. “Photosynthetic algae type the inspiration of the Arctic meals net. Adjustments of their productiveness or species composition can ripple upward to have an effect on fish, seabirds, and marine mammals. Furthermore, photosynthesis performs an necessary position in pure CO2 uptake by the ocean.”
The examine highlights that local weather change within the polar areas does greater than soften ice—it causes elementary shifts in key processes resembling gentle transmission and vitality circulate in marine ecosystems.
The outcomes underscore the significance of incorporating gentle spectra and photosynthesis extra explicitly in local weather fashions and ocean forecasts, particularly in polar areas the place environmental change is accelerating at an unprecedented charge.
Reference: “Lack of sea ice alters gentle spectra for aquatic photosynthesis” by Monika Soja-Woźniak, Tadzio Holtrop, Sander Woutersen, Hendrik Jan van der Woerd, Lars Chresten Lund-Hansen and Jef Huisman, 30 April 2025, Nature Communications. DOI: 10.1038/s41467-025-59386-x
Organizations as we speak face an simple fact: cell units are the trendy gateway to the enterprise. As hybrid work continues and BYOD turns into pervasive, the standard perimeter has all however disappeared. Cell units are actually a major goal within the cyber assault chain — usually exploited first via unpatched vulnerabilities, insecure wifi connections, or Mishing (mobile-targeted phishing). Companies want a extra clever approach to assess threat and management entry, particularly on unmanaged or frivolously managed endpoints.
Massive language fashions like GPT-4 have already develop into a strong software for enterprise. However working by way of public APIs is all the time a threat: information is outsourced, flexibility is proscribed, and prices can shortly escalate.
However there’s a answer — construct your LLM mannequin from scratch. This offers you full management, safety, and customization in your wants. On this information, we’ll present you precisely learn how to do it, with out water and sophisticated phrases.
What’s a Personal LLM?
A non-public LLM (Giant Language Mannequin) is a synthetic intelligence-based system that an organization deploys and makes use of inside its infrastructure: on its servers or in a non-public cloud. Such fashions are utilized in chatbots, search, suggestions evaluation, and different duties involving pure language interplay.
In contrast to public options like ChatGPT, Google Gemini, or Claude, this mannequin solely runs for your corporation and doesn’t share information with exterior providers. That is particularly essential for those who work with private, commercially delicate, or extremely regulated information — for instance, within the monetary, medical, or authorized sectors.
The primary benefit of a non-public LLM is full management over the info, safety, and logic of the mannequin. You’ll be able to customise the system to your business, retrofit it on inner paperwork, and construct it into your merchandise — from chatbots to analytics platforms.
The place are Personal LLMs Utilized?
Personal language fashions are an increasing number of widespread in industries the place safety, accuracy, and information management are significantly essential:
Monetary Expertise (Fintech)
Personal LLMs are used to course of functions, analyze transactions, generate monetary analytics, and help prospects in chat rooms. Such fashions enable for safe processing of non-public and fee information whereas complying with regulatory necessities (e.g., GDPR, PCI DSS).
Drugs and Well being Care
On this space, LLMs assist physicians and employees shortly analyze medical data, generate experiences, confirm appointments, and even predict dangers. All whereas conserving all information in a closed loop, vital for compliance with HIPAA and different medical requirements.
Inside Company Chatbots and Assistants
One of the best a part of LLMs is which you can prepare a non-public language mannequin in your firm’s inner docs, tips, and information base. A wise assistant that offers clear, customized solutions to your staff might help get issues completed sooner and take strain off your help employees.
When Does a Enterprise Want Its LLM?
Generally corporations create their language mannequin not as a result of it’s trendy, however as a result of there is no such thing as a different method. They should adjust to legal guidelines, defend information, and take note of the specifics of the enterprise. That’s why it may be actually essential.
To Comply With Regulatory Necessities (GDPR, HIPAA, and so forth.)
Firms that deal with private information are required to conform strictly with information privateness laws. The usage of public LLMs (akin to ChatGPT or different cloud APIs) might violate GDPR, HIPAA, and different legal guidelines if information is transferred to exterior servers.
Safety of Mental Property and Inside Data
If your organization works with know-how, patent documentation, strategic plans, or R&D information, any leaks could cause severe harm. Coping with a public mannequin that logs or can use your information for additional studying is a threat.
Working with Native or Weakly Structured Information
Many corporations maintain distinctive inner information bases, from technical documentation to company tips. To successfully use them in AI, the mannequin must be additional educated or personalized to the corporate’s specifics. Public fashions don’t enable for this. A proprietary LLM could be educated in your information, together with native recordsdata, information bases, tickets, CRM, and extra.
Help for Extremely Specialised or Non-Commonplace Duties
Off-the-shelf LLMs are good at dealing with normal points, however usually not tailor-made to the terminology and construction of particular industries — be it legislation, building, oil and gasoline, or prescription drugs.
Selecting the Proper Strategy: Construct an LLM from Scratch or Use a Proprietary Mannequin?
When a enterprise decides to create its personal LLM, the following step is to decide on the proper mannequin. There are two important instructions: use open-source options (open-source fashions that may be personalized), or select a proprietary mannequin — an off-the-shelf system from a big know-how firm, akin to OpenAI, Anthropic, or Google.
Each choices can kind the idea of a non-public LLM, however they differ significantly within the diploma of management, value, customization choices, and infrastructure necessities. Beneath, we’ll have a look at the variations between them and the way to decide on an method relying on the enterprise goals.
Well-liked Open-Supply Frameworks
Listed below are essentially the most actively developed and used open-source fashions:
LLaMA (from Meta): a strong and compact structure that’s well-suited for fine-tuning in non-public environments. LLaMA 2 is limitedly licensed, whereas LLaMA 3 is already open supply.
Mistral: quick and environment friendly fashions with excessive accuracy on a small variety of parameters (e.g., 7B). They work particularly properly in era and dialogue duties.
Falcon (from TII): a household of fashions targeted on efficiency and vitality effectivity, appropriate for deployment in enterprise environments.
GPT-NeoX / GPT-J / GPT-2 / GPT-3-like: community-developed fashions with full openness and deep customization.
Comparability of Approaches: Open-Supply vs. Proprietary
To decide on the proper path for personal LLM implementation, there may be worth in understanding how open-source and proprietary fashions differ in key methods, from flexibility and price to safety and compliance. Beneath is a visible comparability of the 2 approaches:
Standards
Open-Supply LLM
Proprietary LLM (GPT-4, Claude, Gemini, and so forth.)
Flexibility
Extraordinarily excessive — mannequin structure could be modified and fine-tuned
Restricted — API doesn’t enable adjustments to inner logic
Information Management
Full management: information by no means leaves the infrastructure
Information is processed on the supplier’s aspect
Prices
Excessive preliminary prices ({hardware}, coaching, upkeep), however cheaper at scale
Low entry value, pay-as-you-go or subscription-based
Safety
Most when deployed regionally
Requires belief within the exterior supplier
Updates & Upkeep
Requires an in-house staff or a technical companion
Dealt with by the supplier — updates, safety, and help included
Regulatory Compliance
Simpler to make sure compliance (e.g., GDPR, HIPAA, NDA, and so forth.)
Tougher to totally comply attributable to exterior information switch
Comparability of approaches: Open-Supply vs. Proprietary
Key Steps to Construct a Personal LLM: From Information to Studying Mannequin
Constructing your personal language mannequin takes each a transparent technique and a step-by-step method. All of it begins with getting your information so as, choosing the proper infrastructure, after which coaching the mannequin so it truly understands and solves actual enterprise challenges.
Dataset Preparation
Step one is working with information. For the mannequin to essentially perceive the specifics of your corporation, it should study from high-quality and clear materials. Because of this all paperwork, texts, and different sources should first be delivered to a standardized format, eliminating duplicates and pointless data.
The information is then partitioned and remodeled right into a construction that the mannequin can perceive. If there may be inadequate data, extra choices are created, for instance, by way of paraphrasing or automated translation. All of that is completed to make sure that the unreal intelligence “speaks” your language and understands the business context.
The information is then divided into coaching, check, and validation information, in order that the mannequin doesn’t simply memorize, however learns.
Organising the Infrastructure
Coaching giant language fashions requires highly effective computing sources: fashionable graphics playing cards, cloud platforms, or in-house servers.
The choice is chosen relying on the extent of safety and availability necessities. If the info is especially delicate, for instance, medical or authorized information, the mannequin could be educated and run inside a closed perimeter, with out Web entry.
It is usually essential to arrange a management system prematurely — monitoring, logs, and backups, in order that every thing works in a steady and clear method.
Mannequin Coaching and Validation
The third step is the precise coaching and validation of the mannequin. This course of requires fine-tuning and fixed high quality management. Specialists choose optimum parameters in order that the mannequin learns sooner and doesn’t lose accuracy.
On the identical time, they consider how properly it copes with the duties at hand: the way it responds, how meaningfully it constructs texts, and whether or not it makes errors. At this stage, you will need to cease coaching in time if the mannequin has reached the specified degree, to be able to keep away from “overtraining”.
Advantageous-Tuning on Inside Information
The ultimate step is making the mannequin actually yours. Even when it’s educated on normal information, it received’t be all that useful till it’s tuned to your organization’s particular content material — issues like inner docs, buyer scripts, information bases, and emails.
This helps the mannequin decide up in your tone, your terminology, and the way your staff truly communicates. You too can use actual worker suggestions to show it what sort of solutions work finest.
Deployment and Integration
As soon as your mannequin is educated and tailor-made to your corporation wants, the following large step is rolling it out the proper method. The way you deploy it performs an enormous position in how steady, safe, and scalable the system shall be as your utilization grows.
Most corporations go together with cloud platforms like AWS, Google Cloud, or Azure — they make it simple to launch, add customers, and push updates with out getting slowed down in complicated setup.
Integration through API and Enterprise Purposes
To allow the mannequin to work together with different digital programs, it’s obligatory to supply it with accessible and dependable interfaces. Probably the most common possibility is REST API. With its assist, LLM could be simply built-in into net functions, company portals, CRM programs, or chatbots.
If excessive responsiveness and minimal latency are a precedence, gRPC is a more sensible choice, particularly when utilizing microservice architectures or embedded in cell functions.
This integration permits the mannequin’s capabilities to be utilized throughout all channels and touchpoints with prospects or workers, making it a full-fledged a part of an organization’s digital infrastructure.
SCAND Use Case: Good Journey Assistant
One of many brightest examples of our follow is the Good Journey Assistant mission developed by the SCAND staff. It is a good cell software wherein a non-public LLM acts as a private assistant for vacationers: it helps plan routes, e book tickets, discover attention-grabbing locations, and kind customized suggestions in actual time.
We additional educated the mannequin on specialised journey information, built-in it with exterior providers — akin to maps, lodge reserving platforms, and airline programs — and deployed the answer on cloud infrastructure for top availability and scalability.
This case research demonstrates how a non-public LLM can develop into the know-how core of a large-scale customized product — dependable, safe, and absolutely personalized for the business.
Challenges and Concerns
Regardless of the excessive worth of personal LLMs, companies face a number of essential challenges when implementing them. To make the mission profitable, these elements needs to be taken into consideration prematurely.
Excessive Computing Necessities
Coaching and deploying language fashions require important sources: highly effective GPUs, subtle structure, and storage programs. It’s important for an organization to know that LLM implementation isn’t just a easy mannequin load, however a full-fledged infrastructure process that requires both funding in its personal servers or using a load-optimized cloud.
Authorized and Moral Dangers
Working with AI in enterprise is more and more regulated by legislation. If you’re processing private, medical, or monetary information, you will need to anticipate compliance with requirements akin to GDPR, HIPAA, and PCI DSS.
Reputational dangers must also be thought-about: the mannequin needs to be designed to keep away from producing discriminatory, deceptive, or malicious responses. These points are solved by way of restrictions, filters, and clear management over what information the AI is educated on.
High quality of Findings and Interpretability
Even a well-trained mannequin could make errors, particularly in new or uncommon conditions. The important thing problem is to make sure that its solutions are verifiable, its conclusions explainable, and that it communicates the boundaries of its competence to the consumer. With out this, the LLM might give the phantasm of confidence when producing inaccurate or fictitious information.
Why Accomplice With an LLM Improvement Firm
SCAND develops language fashions, and dealing with us brings many benefits to companies, particularly for those who plan to implement AI-based options.
To begin with, you instantly get entry to full-cycle specialists: no have to construct a staff from scratch, lease costly tools, and spend months on experiments.
We have already got confirmed approaches to growing and coaching LLMs for particular enterprise duties — from coaching information assortment and transformer structure design to fine-tuning and integration into your IT infrastructure.
Second, it’s threat mitigation. An skilled staff might help keep away from errors associated to safety, scaling, and regulatory compliance.
As well as, we all know learn how to leverage ready-made developments: SCAND already has working options primarily based on generative AI-chatbots for banks, clever journey assistants, and authorized help programs tailored to the required legal guidelines and requirements.
All of those merchandise are constructed utilizing pure language processing methods, making them significantly helpful for duties the place you will need to perceive and course of human language.
Need to implement AI that works for your corporation? We might help.
