In the lead-up to our 9th Annual LDV Vision Summit – the premier global gathering in visual tech – hosted virtually on March 28, we decided to interview some of our past & future speakers.

Meet Luc Vincent of Meta, Steffen Tjerrild of Synthesia, Kristen Fortney of BioAge, Clément Farabet of DeepMind, Camilo Fosco of Memorable, Alexandra Boussommier of ImVitro, Michael Geer of Humanity, Henrik Haugbølle of Uizard, Mariano Battan of Mural, Alexandre Winter of Norbert Health, Andrew Rabinovich of Headroom, Karin Andrea Stephan of Earkick, Adam Paulisick of maad labs, and Rebecca Paoletti of CakeWorks.

Clement Farabet

Clément Farabet is VP of Research at Deepmind. He received his Ph.D. from University Paris-Est while a visiting researcher at NYU then was co-founder & CTO of MADBITS which was sold to Twitter, where he co-founded and led Twitter Cortex for almost 3 years. Before joining Deepmind, Clément was VP of AI & AV Infrastructure at NVIDIA, where he was orchestrating AI training for autonomous vehicles.

What type(s) of visual technologies do you leverage in your business?

Clément: In my past 4 roles (NYU, MadBits, Twitter, NVIDIA), I have built a dozen of visual artificial intelligences. Though the size of these AIs grew considerably in that timeframe (~15y), they are still architecturally similar. The earliest form was made of a convolutional neural network–a basic sequence of convolutions and non linearities like tanh or ReLUs, with only 3-5 layers, capable of basic image segmentation and classification (2010). The most recent form (2022) was made of more complex components, particularly self-attention units, and with more depth (around a dozen), and capable of full blown 3D surround understanding of complex real-world scenes.

The training recipe is still very much the same, all layers are differentiable with respect to their weights and back-propagation is used to compute gradients and update the weights. The key upgrade in the past 10 years since AlexNet was Transformers, which introduced the self-attention unit, which has more powerful learning abilities than previous architectures. One thing that started evolving in the past 10y is also the attention to datasets, and our reliance on synthetically generated datasets to complement real ones.

What visual tech-powered product do you hope exists in 20 years?

Clément: My biggest hope is that we reach a point where we're able to train a transformer-like model in an entirely unsupervised or self-supervised way to emerge a complete and accurate 3D world model, that can be relied on by downstream processing for advanced robotics tasks like autonomous vehicles, robots interacting with objects, etc. And that that training process can require little to no supervision to lead to that 3D representation. Our biggest hurdle today remains how much we lean on supervision/manual labeling to teach these systems.

What advice would you give to a deep tech person who wants to start a business?

Clément: Partner up with an equally talented/amazing product person. The biggest hurdle for a deep tech (or academic) person is to think product instead of technology. An amazing product person, who has a clear product sensibility, understands how to interact with users while leveraging the deep tech's core uniqueness, is what will lead to value creation.

What is the best and the worst personality trait of an investor?

Clément: Best: drive. Worst: ego.

We are honored to host a fireside chat between you and LDV Capital’s Founder and General Partner Evan Nisselson at our 9th annual LDV Vision Summit on March 28th. What are you most looking forward to at our next LDV Vision Summit?

Clément: Catching up with the state of the computer vision industry / startup ecosystem, hopefully learning about new business ideas and applications of the latest technology that we now have in the space.

Kristen Fortney

What type(s) of visual technologies do you leverage in your business?

Kristen: Many of the AI algorithms underlying visual technologies that were initially developed to analyze and interpret images are also broadly applicable to other types of information, including molecular data. At BioAge, our platform applies machine learning to human omics data to identify the molecular patterns that represent successful aging — and we use these discoveries to develop drugs that treat disease by targeting the fundamental mechanisms of aging.

What visual tech-powered product do you hope exists in 20 years?

Kristen: I’d love to see whole-brain connectome mapping — an accurate representation of all of the neuronal connections that make up a human mind. Collection of the data will also require some significant advances in biological imaging, but once the data are harvested, the job of interpreting the data will involve the descendants of today’s visual technologies. And it would be so thrilling to be able to say: here’s the brain, the thing that makes human beings who we are, and this is exactly how it’s put together.

What advice would you give to a deep tech person who wants to start a business?

Kristen: I would say that they’ve picked a very good time to do it. More and more innovation is happening in small, nimble companies, instead of only in academia. A startup can be a great place to pursue a high risk, high reward project that wouldn’t be a good fit for grant-supported research, or even in a larger, more established company.

On March 28th, Kristen will take part in the panel discussion, “How and When Will Deep Tech Increase the Longevity of Humans?”

Luc Vincent

What type(s) of visual technologies do you leverage in your work?

Luc: I am working on AI technology to power Smart Glasses and AR Glasses, a product category that is still in its infancy. These glasses will be loaded with sensors, including IMU, GPS, microphones, and of course cameras. The egocentric data from these sensors - including in particular the visual data - will provide a very rich source of context to trigger experiences, such as shopping, navigation, or media capture.

What visual tech-powered product do you hope exists in 20 years?

Luc: I am looking forward to advanced AR glasses being available in 20 years, and hopefully much sooner. These visual-tech powered glasses will empower and connect people while letting them be more present. They have the potential to be even more useful than smartphones, to billions of people.

What advice would you give to a deep tech person who wants to start a business?

Luc: I'd recommend they consider teaming up with one or two other co-founders, with complementary backgrounds in product management and/or business.

What is the best and the worst personality trait of an investor?

Luc: Best: patience. Worst: impatience.

What are you most looking forward to at our 9th annual LDV Vision Summit, March 28th?

Luc: The whole thing!

Steffen Tjerrild

What type(s) of visual technologies do you leverage in your business?

Steffen: As a visual tech company, we fundamentally leverage a variety of technologies to build our product. This includes computer vision, image and video processing, and graphics rendering.

What visual tech-powered product do you hope exists in 20 years?

Steffen: In 20 years, I hope that there will exist a visual tech-powered product that can seamlessly integrate augmented and virtual reality into our daily lives for practical and recreational purposes.

What advice would you give to a deep tech person who wants to start a business?

Steffen: My advice would be to focus on technological paradigm shifts and how they apply to big industries and problems and then assemble a diverse team of ambitious, smart and humble people.

What is the best and the worst personality trait of an investor?

Steffen: Best: adaptable. Worst: arrogant.

Thrilled to have you speak at our 9th annual LDV Vision Summit about your generative AI video platform on March 28th. What are you most looking forward to?

Steffen: Looking forward to exploring the latest advancements in visual technology and networking with like-minded operators and investors in the space.

Camilo Fosco

What type(s) of visual technologies do you leverage in your business?

Camilo: We mostly use deep neural networks to analyze and make predictions over images and videos. Our networks take visual stimuli as inputs and estimate rich compressed representations (embeddings) that can be used in multiple downstream tasks, such as predicting recall probabilities, estimating distributions of attention, detecting objects and actions, etc. Lately, we've also been using generative models to allow our clients to create ads from simple text prompts, with an emphasis on creative effectiveness: our generations are guided by our branding and performance models, ensuring that generations maximize KPIs of interest.

What visual tech-powered product do you hope exists in 20 years?

Camilo: True, fully-functioning augmented reality, embedded in our eyes or relayed through comfortable contact lenses. Hopefully in 5-10 years, a world generator: a fully explorable visual world generated on the fly according to the user's actions.

What advice would you give to a deep tech person who wants to start a business?

Camilo: Build fast, but reusable. Start thinking about scaling early or it will catch up with you. Be smart when hiring. The choice that solves for urgency rarely solves for importance.

What is the best and the worst personality trait of an investor?

Camilo: Best: insightfulness. Worst: indifference.

Excited to have you speak about generative AI & cognitive science to make ads memorable at our 9th annual LDV Vision Summit on March 28th. What are you most looking forward to?

Camilo: Hearing about the latest tech being developed in the LDV community, and getting to know bright people interested in visual tech!

Alexandra Boussommier

What type(s) of visual technologies do you leverage in your business?

Alexandra: Computer vision that analyzes videos of embryos as they develop in the lab to objectively decipher their potential to lead to pregnancies. The algorithms have been trained to freely focus on any information they recognize as important in these complex videos, without biasing them with human-defined criteria to ensure we can complement and empower doctors in their decision-making. This visual information is then also analyzed with machine learning with a myriad of patient's characteristics to get closer to the multi-factorial clinical reality that complicates medical decisions. In addition to providing a new AI-powered datapoint, our computer vision also extracts automatically key cellular events normally identified by doctors, to automate embryologist's workflow.

What visual tech-powered product do you hope exists in 20 years?

Alexandra: Multi-scale non-invasive microscopy to be able to routinely understand, monitor and treat diseases from the microscopic cellular level to the macroscopic tissue and body level.

What advice would you give to a deep tech person who wants to start a business?

Alexandra: Do not push your technology. Focus all your energy first on finding a problem that pulls it.

What is the best and the worst personality trait of an investor?

Alexandra: Worse: overbearing. Best: (real) risk-taker.

Excited to have you speak about empowering IVF clinics to increase workflow and maximize pregnancy success rates with AI at our 9th annual LDV Vision Summit on March 28th. What are you most looking forward to?

Alexandra: Identifying synergies between successfully commercialized computer vision tools across fields and continents, and meeting optimistic computer vision founders and investors to keep track of the field's expectations in the coming years.

Michael Geer

What type(s) of visual technologies do you leverage in your business?

Michael: At Humanity, our mission is to add a billion healthy years to everyone before 2030, and this means we need to leverage health data biomarkers that almost everyone can access. Our app therefore leverages the rich data coming from phones and wearables, which in turn heavily leverage visual sensors. These are the unsung building blocks of health equity.

What visual tech-powered product do you hope exists in 20 years?

Michael: As you know, the different visual spectrum sensors on wearables are rapidly replacing the need for needles and other invasive procedures to measure the body. We expect visual sensors to one by one replace traditional blood test methods over the next 5 years. Either by way of monitoring the person's vein (as will be the case with glucose) or by allowing for finger prick blood droplet analysis right then and there in the person's home.

When we combine that rich and low cost flow of longitudinal data on each of us, the power of AI deep learning model training, and the ease of interaction now being demonstrated with LLMs (Large Language Models - like ChatGPT), we have the ability to not only predict a person's future health, but to expertly guide them to do exactly what will make them healthier for decades longer.

What advice would you give to a deep tech person who wants to start a business?

Michael: Given all the above, it is clear there is no need for any of us to settle for small incremental change. Make sure the ARR of your portfolio or company is the least exciting thing about what you are doing.

You will take part in the panel discussion on longevity tech at our 9th annual LDV Vision Summit on March 28th. What are you most looking forward to?

Michael: One look at the speaker list of LDV Vision Summit this year, makes me even more sure that the future is very bright indeed.

Henrik Haugbølle

What type(s) of visual technologies do you leverage in your business?

Henrik: We’re leveraging computer vision, generative AI, deep learning and machine learning, supported by heuristics and natural language processing.

What visual tech-powered product do you hope exists in 20 years?

Henrik: I dream of a paradigm shift in hardware, getting rid of screens, keyboards, and trackpads as we know them today, with the aim of reducing eye and muscle strain. I believe there’s still lots of avenues to be explored in this field powered by computer vision and AI.

What advice would you give to a deep tech person who wants to start a business?

Henrik: Validate early and often, take shortcuts and be scrappy, but also know when to sink time into developing AI moat – be aware that the latter will take more data, time, and resources than you anticipate, making the early validation even more important.

Productivize the tech if possible. A deep tech feature can be exciting in itself, but a product will build a business.

What is the best and the worst personality trait of an investor? (in one-word answers for each)

Henrik: Best trait: ambitious realism. Worst trait: narrow-minded.

We can’t wait to hear your keynote speech, “The First Generative AI Design Platform To Auto-Generate End-To-End Design Via Text” at our 9th annual LDV Vision Summit on March 28th. What are you most looking forward to?

Henrik: Showcasing the next generation AI in software design and networking with like-minded pioneers within vision tech.

Karin Andrea Stephan

What type(s) of visual technologies do you leverage in your business?

Karin Andrea: Earkick is the first data-driven and real-time mental health measurement platform that enables companies to measure the mental health of their workforce via multimodal sentiment analysis of physiological biomarkers while maintaining the highest level of privacy. Health signals are captured via voice, text and video. We leverage visual sentiment analysis and facial biomarker detection to estimate the user’s mental health.

What visual tech-powered product do you hope exists in 20 years?

Karin Andrea: I strongly believe that 20 years from now, people will have a virtual personal companion that is present via XR and can interact with the user. For this, the virtual companion has not only "see" the user, but also see and understand context and environment. That way this virtual companion can help the user to accomplish various tasks in the best way possible.

What advice would you give to a deep tech person who wants to start a business?

Karin Andrea: Starting a tech business is pretty hard and starting a deep tech business is even harder as it often relies on data that first needs to be collected. This can be a huge challenge. At Earkick, for example, we first had to develop a data-collection solution that could be used by thousands of users in order to generate enough data. This enables us to focus on our intended product to help people manage their mental health. I would advise people to leverage existing solutions in order to most efficiently and effectively validate product market fit.

What is the best and the worst personality trait of an investor?

Karin Andrea: Best: no-bullshit; Worst: bullshit.

What are you most looking forward to at our 9th annual LDV Vision Summit, March 28th?

Karin Andrea: Hearing from and meeting amazing, smart people who are all pushing the limits on what is possible in tech.

Mariano Battan

What type(s) of visual technologies do you leverage in your business?

Mariano: People's eyes and hands. Anybody can sketch and that helps explain things much better than words sometimes. Look forward to more AI-assisted diagramming coming.

What visual tech-powered product do you hope exists in 20 years?

Mariano: It exists... but doesn't go all the way. Scan a LEGO pile and sort by the different builds that were destroyed in there.

What advice would you give to a deep tech person who wants to start a business?

Mariano: The same I give any entrepreneur: focus on a particular segment and use case first. So much easier to build product, brand and revenue than going horizontal and "platformy" to start with. Then you can go to other segments/use cases or make it a platform.

What is the best and the worst personality trait of an investor?

Mariano: Best: curious. Worst: stubborn.

We are honored to host a fireside chat between you and LDV Capital’s Founder and General Partner Evan Nisselson at our 9th annual LDV Vision Summit on March 28th.

What are you most looking forward to?

Mariano: Inspiration! Besides the joke about the LEGO vacuum cleaner above, I'm delighted by all the automagical tools out there. I think that there is an opportunity to stitch a lot together to solve for people's lives and reduce time to do certain boring things.

Alexandre Winter

What type(s) of visual technologies do you leverage in your business?

Alexandre: We use multispectral, calibrated sensors – long wave infrared, visible light and FMCW radar imaging. We use ML models and analytical models to combine them efficiently. These sensors capture very different aspects of the world, of humans, and offer amazing possibilities.

What visual tech-powered product do you hope exists in 20 years?

Alexandre: Bicycle, scooter but also power tool ADAS or safety systems (no more horror movies with chainsaws), drones flying indoors autonomously (without breaking anything), accurate subterranean and through wall imaging (visualize the pipes, cables, or anything else underground or behind walls).

What advice would you give to a deep tech person who wants to start a business?

Alexandre: Focus your efforts on solving one hard problem – not many, just one. Make all the rest as simple as possible. Always be resilient – it is not going to work the first time but you will find a way to make it work.

What is the best and the worst personality trait of an investor?

Alexandre: Best: anxious. Worst: optimistic.

What are you most looking forward to at our 9th annual LDV Vision Summit on March 28th?

Alexandre: Reconnecting with the unique community that was created around this event, and seeing a lot of new exciting sensing technologies.

Rebecca Paoletti

What type(s) of visual technologies do you leverage in your business?

Rebecca: As a video agency, we are constantly looking for ways to make video production, distribution and monetization more efficient, and more impactful. Companies that support multi-user video editing and approval work-flow, easy archiving, tagging and metadata capture, searchability and discoverability and shoppable elements are all critical to the work we do on behalf of our clients. We were early users of AI to both streamline productivity and create more engaging video experiences, and continue to look for next-generation implementations that can bring new sparks of innovation to video viewing.

What visual tech-powered product do you hope exists in 20 years?

Rebecca: Fully immersive virtual reality without a clunky headset. And, Minority Report-styled billboards and messaging that are customized to – and changeable by – me.

What advice would you give to a deep tech person who wants to start a business?

Rebecca: Find a partner who can help simplify your ideas to your future customer, and be sure that there is actually a consumer need for the idea that you have.

What is the best and the worst personality trait of an investor?

Rebecca: Best – candor. Worst – bullying based on perceived power.

What are you most looking forward to at our 9th Annual LDV Vision Summit?

Rebecca: The imagination, innovation, networking, community and fierce activity that is a part of every LDV Vision Summit!

Andrew Rabinovich

What type(s) of visual technologies do you leverage in your business?

Andrew: At Headroom, we take a multimodal approach to artificial intelligence. This includes analyzing speech, language, and vision. During video communication and collaboration, non verbal cues are often lost, and communication becomes challenging and ineffective. Headroom is the only video conferencing platform that interprets non verbal cues in real time. These include hand and body pose tracking, hand and head gesture recognition, gaze tracking and affective computing. Also, to improve the visual fidelity of the experience, we apply scene understanding and super resolution techniques.

What visual tech-powered product do you hope exists in 20 years?

Andrew: I am a big believer in head mounted displays and mixed reality. Virtual content must coexist with the physical world and be ubiquitously available to all. I look forward to the development of practical head mounted displays to deliver such a natural and intuitive experience to interact with the world.

What advice would you give to a deep tech person who wants to start a business?

Andrew: Remember that a fascinating technology isn't the end goal, solving a real problem is. Dream big and go for it.

What is the best and the worst personality trait of an investor?

Andrew: desire to help / patience.

Adam Paulisick

What type(s) of visual technologies do you leverage in your business?

Adam: Like many of us, I wear lots of hats these days around trying to have an impact on the world and whether it’s lecturing at Carnegie Mellon University, instigating with maad labs, building my own products or services, or promoting entrepreneurship in my local community I see visual tech at every turn.

For example, TheShop.org (a local Pittsburgh community workspace founded to support inclusive entrepreneurship) benefits from flexible drone footage but more recently an AR-powered digital workforce (meet Rafiyq, the actual new owner of The Shop - Homewood and CEO of PriveMgt.com) – click the URL and keep hitting “allow” to experience how we offer tours of the new space no matter where you are.

What visual tech-powered product do you hope exists in 20 years?

Adam: My wish list is long but I hope in (less than) 20 years I can dispatch a series of devices to go and scout a vacation spot, or restaurant, or to ideally be a digital shopper (online or in a brick and mortar environments) that would help extend my physical reach without needing to fully interpret experiences by being present e.g., connecting me visually with the place or space with a local robot that I could rent for just a day or two. Something that reminds me of how I used to experiment with early concepts of this idea at BCG where we had remote attendance via robots.

**Adam took part in the panel discussion “How do marketers and publishers successfully track and increase video revenue?” at our 1st annual LDV Vision Summit back in 2014. ©Dan Taylor**

What advice would you give to a deep tech person who wants to start a business?

Adam: Remind yourself that your largest competitors are largely “do nothing” or “I don’t see enough value to change behavior” far ahead of any direct or even indirect competitors.

What is the best and the worst personality trait of an investor?

Adam: Worst: helicopter parents – come only when things are externally great or terrible and never help with the messy middle of scaling . Best: commercial - investors who know that all the product advice or feelings are a fraction as valuable as qualified sales introductions.

What are you most looking forward to at our 9th annual LDV Vision Summit?

Adam: I can’t wait to hear the speakers share lived experiences and in particular Reshma Sohoni from Seedcamp who always brings living and breathing examples of the MAYA principle (Most Advanced, Yet Acceptable) when talking about bleeding-edge or leading-edge technologies.

LDV Blog

14 Past and Future LDV Vision Summit Speakers Share Their Views On What’s The Next Big Thing in Visual Tech

Clement Farabet

Kristen Fortney

Luc Vincent

Steffen Tjerrild

Camilo Fosco

Alexandra Boussommier

Michael Geer

Henrik Haugbølle

Karin Andrea Stephan

Mariano Battan

Alexandre Winter

Rebecca Paoletti

Andrew Rabinovich

Adam Paulisick