Topics
recent
AI
Amazon
Image Credits:Juna.ai
Apps
Biotech & Health
Climate
Image Credits:Juna.ai
Cloud Computing
Commerce
Crypto
Example Juna.ai dashboard.Image Credits:Juna.ai
enterprisingness
EVs
Fintech
Juna.ai copilot.Image Credits:Juna.ai
Fundraising
Gadgets
stake
Government & Policy
ironware
Layoffs
Media & Entertainment
Meta
Microsoft
Privacy
Robotics
Security
Social
Space
Startups
TikTok
Transportation
Venture
More from TechCrunch
Events
Startup Battlefield
StrictlyVC
Podcasts
Videos
Partner Content
TechCrunch Brand Studio
Crunchboard
touch Us
AI agents are all the rage , a tendency drive by the productive AI and large nomenclature theoretical account ( LLM ) roaring these past few year . get people to agree onwhat exactly AI federal agent areis a challenge , but most contend they are software programs that can be assign tasks and given determination to make — with varying degree of liberty .
In short , AI agents go beyond what a simple chatbot can do : They help people get things done .
It ’s still former days , but thelikes of Salesforceand Google are alreadyinvesting heavy in AI agents . Amazon CEO Andy Jassyrecently hintedat a more “ agentic ” Alexa in the futurity , one that ’s as much about natural process as it is wrangle .
In tandem , startupsare also raisingcash off the hype . The latest of these is German companyJuna.ai , which want to assist factories be more effective by automatise complex industrial unconscious process to “ maximize production throughput , increase zip efficiency and deoxidize overall discharge . ”
And to commit that off , the Berlin - based inauguration today state that it has kick upstairs $ 7.5 million in a seed round from Silicon Valley speculation uppercase firmKleiner Perkins , Sweden - basedNorrsken VC , and Kleiner Perkins ’ chairmanJohn Doerr .
Self-learning is the way
Founded in 2023 , Juna.ai is the handiwork ofMatthias Auf der Mauer(pictured above , left ) andChristian Hardenberg(pictured above , right ) . Der Mauer previously founded a predictive machine maintenance startup called AiSight andsold it to Swiss smart sensor company Sensirionin 2021 , while Hardenberg is the former chief engineering science officer atEuropean food delivery giantDelivery Hero .
At its core , Juna.ai wants to help manufacturing facility transform into smarter , self - learning systems that can deliver undecomposed margins and , ultimately , a abject carbon copy footprint . The company pore on “ heavy industries ” — industries such as steel , cementum , newspaper , chemical , wood and textile with large - scale yield processes that eat slews of raw materials .
Join us at TechCrunch Sessions: AI
Exhibit at TechCrunch Sessions: AI
“ We work with very process - driven industries , and it mostly involve habit cases that use a lot of energy , ” Auf der Mauer enjoin TechCrunch . “ So , for exemplar , chemic reactors that habituate a lot of heat in ordering to bring forth something . ”
Juna.ai ’s software system integrates with manufacturers ’ production tool , like industrial software system fromAvevaorSAP , and looks at all its historical information garnered from motorcar detector . This might involve temperate , pressure , velocity , and all the measurements of the given outturn , such as quality , thickness , and color .
Using this selective information , Juna.ai avail company school their in - house agents to figure out the optimal configurations for machinery , giving manipulator real - clock time data and guidance to assure everything is running at peak efficiency with minimum wastefulness .
For example , a chemic plant that produces a limited kind of carbon copy might practice a nuclear reactor to mix different oil together and put it through an energy - intensive burning mental process . To maximize the production and minimize residual waste , conditions need to be optimum , admit the layer of gases and oils used , and the temperature applied to the unconscious process . Using historical data to establish the ideal stage setting and taking real - fourth dimension conditions into story , Juna.ai ’s agentive role purportedly recite the wheeler dealer what changes they should be take to achieve the best yield .
If Juna.ai can help companies okay - tune their product equipment , they can amend their throughput while reducing energy economic consumption . It ’s a win - winnings , both for the customer ’s bottom tune and its atomic number 6 footprint .
Juna.ai says it has built its own custom AI models , using open germ shaft such asTensorFlowandPyTorch . And to train its models , the company usesreinforcement learning , a subset of machine learning ( ML ) that involve a model learning through its interactions with its environment — it tries different action , observes what happens , and better .
“ The interesting thing about strengthener scholarship is that it ’s something that can take action , ” Hardenberg told TechCrunch . “ Typical models only do foretelling , or maybe generate something . But they ca n’t manipulate . ”
Much of what Juna.ai is doing at present is more kindred to a “ co-pilot ” — it serves up a sieve that tells the operator what fine-tune they should be making to the control . However , many industrial processes are fantastically insistent , which is why enabling a system to take real actions is helpful . A chill organisation , for example , might require changeless mulct - tuning to see to it a simple machine maintains the right temperature .
manufactory are already well wonted to automating system controls usingPIDandMPCcontrollers , so this is something that Juna.ai could practicably do , too . Still , for a fledgling AI startup , it ’s easier to sell a co-pilot — it ’s baby step for now .
“ It ’s technically possible for us to rent it fly the coop autonomously decent now ; we would just need to follow out the connection . But in the end , it ’s really all about building trustfulness with the client , ” Auf der Mauer say .
Hardenberg added that the benefit of the startup ’s weapons platform does n’t lie in saving labor , note that factories are already “ quite effective ” in term of automating manual processes . It ’s all about optimise those processes to cut costly waste .
“ There ’s not a lot to gain by removing one person , compared to a appendage that cost you $ 20 million in vim , ” he articulate . “ So the veridical gain is , can we go from $ 20 million in vitality to $ 18 million or $ 17 million ? ”
Pretrained agents
For now , Juna.ai ’s great promise is an AI broker tailored to each customer using their diachronic data . But in the future , the company plans to offer off - the - shelf “ pretrained ” agents that do n’t require much in the fashion of training on a new client ’s data .
“ If we build model again and again , we get to a stead where we can potentially have pretending templates that can be reused , ” Auf der Mauer said .
So if two ship’s company apply the same kind of chemic nuclear reactor , for instance , it might be potential to lift - and - break AI agents between client . One manikin for one machine , is the general nub .
However , there ’s no ignoring the fact that enterprises have been hesitant to dive headfirst into the burgeon AI rotation due to data privacy business . These headache are n’t lost on Juna.ai , but Hardenberg say that it has n’t been a major issue so far , partly due to its data residency controls , and partly due to the hope it gives customers in term of unlocking latent value from huge banks of data .
“ I was seeing that as a potential trouble , but so far , it has n’t been such a big problem because we leave all data in Germany for our German customer , ” Hardenberg aver . “ They get their own waiter do up , and we have top - notch security guarantee . From their side , they have all this data lie around , but they have n’t been so effective at create value from it ; it was mostly used for alerting , or possibly some manual analytics . But our view is that we can do much more with this data — build up an intelligent factory , and become the brain of that factory based on the data they have . ”
A piffling more than a year since its innovation , Juna.ai has a fistful of customers already , though Auf der Mauer said he ’s not at liberty to unveil any specific names yet . They are all base in Germany , though , and they all either have subsidiaries elsewhere or are subsidiaries of ship’s company establish elsewhere .
“ We ’re planning to grow with them — it ’s a very expert elbow room to expand with your customers , ” Hardenberg added .
With the sweet $ 7.5 million in the bank , Juna.ai is now well - financed to expand beyond its current headcount of six , with plan to double down on its technical expertise .
“ It ’s a software company at the end of the Clarence Shepard Day Jr. , and that basically means people , ” Hardenberg pronounce .
