I’ve written before about how important it is to nurture innovation in-house. Businesses run at least partly on ideas, and while it might be easier to go out and buy successful innovators, it’s much more expensive than encouraging innovation to grow in-house.
Innovation and applying an innovative spirit at work has never been more important. AI is in many ways a resource we’re all still working to harness and shape, which is why I believe so many of the successful stories we’re hearing come from bottom-up implementations. That is, people doing it themselves, on their own initiative.
This includes creative workers, students, or young explorers in the wild.
Today I draw from a few publications we love to share profiles of some of these breakthroughs and the innovators behind them, including MIT Technology Review’s 35 Innovators Under 35 2025, the Forbes 30 Under 30 series, and the Time100 AI 2025.
I hope you find these breakthroughs as exciting and encouraging as I do.
New AI Innovations from Young Tech Innovators in 2025
Let’s start where you’d expect to see some of the most exciting work: novel AI. The innovators below are working to fix problems with early systems and bring the technology in exciting new directions.
Hallucination Resistance for Critical Tasks, Better Quality Research, and Understanding Failures
Hallucinations in AI have been sticky and though rates come down, they continue to bog down businesses who worry about trust.
New Carnegie Mellon Language Technologies Institute faculty member Akari Asai is recognized on MIT’s list for her work in retrieval-augmented generation (RAG), specifically being part of creating the Self-RAG framework which achieves a notable step forward in reducing hallucinations from LLMs. Their work uses augmented language models (LMs) that pair together for tools, retrieval, and self-checking. Her work goes in contrast to the main drive of continued scaling.
Ensuring trust with tech requires understanding, and the black box nature of GenAI systems makes this a challenge. Google DeepMind’s Neel Nanda takes this on in his work. He says:
“I see my job as: Do research such that by the time we make human-level AI, it is safe and good for the world.”
He works in mechanistic interpretability (aka mech interp, aka using math tools to learn what AI is doing). His work enables better debugging, guardrails, and transparency on realistic limits.
Recognized by both MIT and Time for 2025, former Google researcher and current Samaya AI Co-Founder and CEO Maithra Raghu has taken on finance reliability. Like our first example, her in-house models are smaller and more specialized, delivering more accurate, domain-specific insights.
She believes deeply in specialization, or AI experts made for specific use cases.
Another former Google employee (singled out by Forbes), Aidan Gomez is already known for coauthoring the famous AI paper on transformers called Attention Is All You Need.
He’s since launched Cohere, which helps companies implement AI-driven products and has massive investments from companies like Nvidia and Oracle. They focus on AI for regulated industries and integrating RAG solutions deeply.
AI’s Cross-Modality and Real-World Know-How
We take for granted that AI can work across text, audio, images, and video. That it can one of these in for prompt and answer out in another.
But someone had to bring this to reality, and one of the key players in that was OpenAI’s Mark Chen.
His work helped adapt transformers for images (“like a strange language”), with pixels as tokens. More recently he’s worked on their Codex coding product, working to keep models following user intent.
As a Chicagoan, I love seeing the innovations that spark at Northwestern University, and MIT’s 35 Under 35 for 2025 recognizes NW Computer Science Assistant Professor Manling Li.
One of the commonly touted limitations of LLMs is their lack of understanding, and that’s something she takes on in her research.
She works to help systems connect goings-on in the real world, to figure out relationships and even causality. Her frameworks have already been adopted by DARPA in the US government and have also been shared as open-source tools.
And speaking of understanding, helping spur robotics breakthroughs and better prosthetics and training methods is the goal of crafting a unified computer model for human movement.
MIT’s Nidhi Seethapathi has done this with a focus on human stability and movement efficiency, but also how it ties into energy costs.
Climate and Biotech Innovations from Rising Stars
AI is currently quite energy-intensive, but already we see the technology also being used to help improve energy inefficiencies in other processes. Some are profiled below.
Making More from Waste in Sustainable Tech Startups
One example of an innovation seeking to reduce our carbon footprint is pioneered by Calcarea Co-Founder and Chief Technology Officer Pierre Forin.
Tackling maritime shipping’s pollution cost has long been challenging, as massive ships require energy to carry them thousands of miles overseas.
His innovation captures carbon dioxide and filters it through seawater and then limestone, creating water that can be released harmlessly along the trip. This avoids the big costs normally associated with such a capture process. And they’re now testing his tech for use on land, as for cement plants near coastlines.
Speaking of water, the fashion industry uses more water than any other sector but agriculture. It’s also responsible for some 8% of global greenhouse gas emissions.
Now with the help of AI and following the behavior of trees, startup Rubi, founded by twins Neeka and Leila Mashouf, is converting carbon dioxide directly into cellulose that can then be used in fabrics.
A similar breakthrough is being pioneered by Dioxycle, a Paris startup co-founded by Sarah Lamaison. She left Stanford in 2021 to launch the company, and their technology converts carbon dioxide and carbon monoxide into ethylene, which can be used to create plastic and textiles without the usual carbon dioxide cost. It also can match, or even beat, the current cost of creating ethylene using fossil fuels.
Moving from pollution to weather, UCLA Assistant Professor of Computer Science Aditya Grover helped build the first foundation AI model for weather and climate.
With his academic associates and researchers at Microsoft, he’s trained AI on massive datasets from real-world observation, developing open-source tools for adaptable weather and climate predictions.
These are already helping improve forecasts, with the goal of reducing disaster damage and even giving better insights on crop yields.
AI-Powered Healthcare Startups: Invisible Mice, Faster Sequencing, and Smart Materials
One of my favorites of these innovations has to be making mice transparent.
Yes, you read that correctly.
The University of Texas at Dallas’s Zihao Ou’s research, which has been published in Science, uses chemicals commonly found in food dyes to peer inside bodies without surgery.
Their approach, which combines these molecules with the right kind of light, can temporarily make layers of a living mouse—like its skin, muscle, and tissue—transparent, revealing the working organs inside.
They are working now to repeat this process on human skin, which is thicker, but believe it may be possible for use on patients within five or 10 years.
MIT’s 2025 Innovator of the Year is Princeton’s Sneha Goenka, who used AI to reduce diagnostic times in sequencing from weeks to under eight hours. Her system has already saved lives.
It began five years ago, when her team devised new hardware architectures and improved computations to speed up each stage of the genome sequencing process.
Their innovations have been dramatic enough to move the process from research to life-saving medicine.
Merging medicine with wearable technology innovations is the aim of Irmandy Wicaksono, a researcher from the National University of Singapore.
His textiles are able to sense aspects of a wearer’s health, from respiration to heart rate to posture. And all in real-time. They’re unique because the sensors are imbedded in the yarns of the fabric itself.
And best yet, they’re highly wearable, and even washable. Wicaksono has also created a yoga mat that infers a user’s movement with 99% accuracy.
The Forbes, Time, and MIT Innovators Under 35 Working for Global Good
In addition to pushing technology into new areas, there’s also innovation aimed at securing existing ones. And here we look at snapshots of young people helping make organizations and individuals safer in our rapidly changing world.
Safety and Verification of Personhood
About five months ago, I posted a CNN video segment on LinkedIn about the World Network Orb. Built by an organization that was co-founded by OpenAI’s Sam Altman (now called World ID), it seeks to fix an increasingly challenging problem: proof of human.
The Orb is hardware that scans your iris and gives you a unique ID which can be used as a de-identified passport for the internet. Reddit is currently exploring a partnership to fix their spam problem while preserving anonymity, and Tinder is trialing it for an anti-catfish offering.
World ID CEO Alex Blania is being recognized in the Time100 AI 2025 for these efforts, and he sees their work as essential for maintaining trust with the rise of deepfakes.
AI Now’s Chief AI Scientist Dr. Heidy Khlaaf takes on AI safety, and her work started at OpenAI on Codex. With a background in nuclear and automotive safety, she developed a methodology for AI that’s been adopted at AI labs globally.
And now she’s taking on AI use in military applications, calling for nations to stop using commercially available models and instead work to ensure higher dependability rates.
For Regulations and Transparency
AI regulation is one concern, but AI is also being used to aid in regulatory compliance.
Forbes’ 30 Under 30 details one such innovator, Georgina Steele, whose UK startup Maiven uses AI to combine company policy with their data to find potential compliance issues before they result in massive fines.
AI’s capacity to wrangle data at scale makes it appealing for use in the public sector, and there are numerous examples of the benefits that come from this. But there are also risks and potential biases, which are being taken on by French researcher Soizic Pénicaud.
She’s built a public repository of algorithms in use by the French government (with an informal group of representatives from 15 other countries), to ensure transparency and accountability.
Pénicaud believes in AI’s ability to efficiently allocate energy, for example, but also that transparency and trust are key, and she’s working to ensure systems function without public harm.
Voice AI, Customer Service, and Digital Twins Technology
Voice AI has moved from video games and novelty uses to serious business applications.
Among the young stars in this field is Time100’s Mati Staniszewski. The ElevenLabs co-founder and CEO is driving a mission to make AI speak every language, dub and clone easily, and be seamlessly conversational for creative use as well as in enterprise.
Their work helped Virgina’s congresswoman Jennifer Wexton address the House of Representatives despite having lost her voice due to a neurological condition.
They’re also part of a rise in solutions that give companies 24/7 global support and outreach that’s far more user-friendly than traditional systems.
In an adjacent field, Forbes recognizes pioneer Caoimhe Murphy, a co-founder of English startup Anam and former member of Synthesia. Anam’s emphasis is AI avatars that handle customer service and sales, and they already have 2,000 clients across sectors like education, sales, customer service, and healthcare.
Their emphasis, like Staniszewski’s, is a new way of interacting with machines that’s more conversational and recognizable.
Conclusion: Finding New Ways to Use Technology Is Only the Beginning
I’ve picked out some of the innovations and young innovators that struck me, but one of the most exciting things about AI is how it can empower young and old alike.
In 2023, the average age of Fortune 500 CEOs was just under 58. Does that make today’s leaders less ready to seize on AI in their own unique ways?
The answer, of course, is no. But for those with established workflows, who already enjoy success in a given field, there can be less motivation to fully explore AI and other new technologies. Sometimes the drive can be more from a fear of being outdistanced or left behind.
Looking at bold, original applications of technology to solve existing problems in new ways can be a refreshing reminder of the joy and power in exploration and discovery.
References
30 Under 30 Europe Technology 2025: The Young Innovators Shaping The AI Revolution At Breakneck Speed, Forbes
35 Innovators Under 35 2025, MIT Technology Review
Time100 AI 2025, Time
2025 Innovator of the Year: Sneha Goenka for developing an ultra-fast sequencing technology, MIT Technology Review
Meet the typical Fortune 500 CEO: A total Gen Xer. Basically Keanu Reeves, Fortune
