Inference at Scale: Building the Infrastructure for AI’s Next Phase

Riley Finch is an Investment Associate at NextGen Venture Partners (NGVP), where he sources early-stage investments and works with the firm's network of more than 1,700 venture partners to support founders through sourcing, diligence, sales, hiring, and fundraising. ~~~~~~~~~~~~~This episode is brought to you by Nebius — the ultimate cloud for AI innovators.Nebius provides AI infrastructure you can count on, combining reliability and speed with flexibility and engineering support unmatched by hyperscalers.AI leaders like Meta, Shopify, and Higgsfield already partner with Nebius to run their AI workloads. Plus, venture-backed startups can save up to $150,000 on compute costs when they apply for access. Visit nebius.com or nebius.com/startups to learn more~~~~~~~~~~~~~NGVP typically invests $500,000 to $2 million in Pre-Seed and Seed-stage companies. Before joining NGVP, Riley was an Investor at 1Sharpe Ventures and led operations and platform initiatives at Afore Capital. He previously held go-to-market and investment roles at fintech companies including Clearco and Wealthsimple. Outside venture capital, Riley spent four years on the Canadian National Sailing Team, competing toward the Olympic Games. He holds a BBA in Finance from Wilfrid Laurier University.TopicsWhat makes exceptional founders and early-stage companies stand out at the Pre-Seed and Seed stagesHow venture partner networks create value for startups beyond capital Applying an operator's perspective to venture investing, from fintech go-to-market experience to founder support#VentureCapital #EarlyStageInvesting #Startups #Fintech #FounderSupport

Kevin B. Boyd is the Associate Vice President and Chief Information Officer at the University of Chicago, where he leads Information Technology Services, overseeing the university’s core IT infrastructure, cybersecurity, enterprise data systems, research computing, and digital learning technologies. ~~~~~~~~~~~~~This episode is brought to you by Nebius — the ultimate cloud for AI innovators.Nebius provides AI infrastructure you can count on, combining reliability and speed with flexibility and engineering support unmatched by hyperscalers.AI leaders like Meta, Shopify, and Higgsfield already partner with Nebius to run their AI workloads. Plus, venture-backed startups can save up to $150,000 on compute costs when they apply for access. Visit nebius.com or nebius.com/startups to learn more~~~~~~~~~~~~~With expertise in enterprise transformation, AI, and data strategy, he focuses on building high-performing technology organizations and aligning IT investments with institutional priorities. He is also a board and advisory leader and an ORBIE CIO of the Year.TopicsIntegrating AI and data strategy into enterprise technology and organizational transformationBalancing cybersecurity, infrastructure modernization, and emerging technology adoptionBuilding high-performing IT teams and aligning technology investments with institutional goals#EnterpriseAI #CIO #DigitalTransformation #DataStrategy #Cybersecurity

Jay Zhao is the Founder and Managing Partner of Leonis Capital, a research-driven venture firm investing in pre-seed and seed-stage AI-native, fintech, and enterprise software companies. ~~~~~~~~~~~~~This episode is brought to you by Nebius — the ultimate cloud for AI innovators.Nebius provides AI infrastructure you can count on, combining reliability and speed with flexibility and engineering support unmatched by hyperscalers.AI leaders like Meta, Shopify, and Higgsfield already partner with Nebius to run their AI workloads. Plus, venture-backed startups can save up to $150,000 on compute costs when they apply for access. Visit nebius.com or nebius.com/startups to learn more~~~~~~~~~~~~~Founded in 2021, Leonis backs technical founders through original research into emerging technologies and the markets they could transform. Across more than 14 years in venture capital, Jay has backed 11 companies that reached unicorn status, including three that went public. His investments include Marqeta, Anaplan, Lime, MaintainX, and Spectro Cloud. Before founding Leonis, he co-founded T Fund and held senior investment roles at Walden Venture Capital and Granite Ventures. His work has been featured in CNBC and The Wall Street Journal, and he has spoken at Harvard Business School on research-driven AI investing.TopicsUsing original technology and market research to identify emerging AI investment opportunities Evaluating technical founders and early-stage companies across AI, fintech, and enterprise softwareLessons from 14 years of venture investing, including backing companies from their earliest stages through IPOs and major exits#VentureCapital #AIInvesting #ArtificialIntelligence #EarlyStageInvesting #Startups

As AI moves from training models to deploying them at massive scale, inference is emerging as one of the defining infrastructure challenges of the AI era.At Smart AI 100 2026, Grace Gong, Founder and CEO of Smart Venture Media, moderated a conversation with investors, founders, and operators shaping the future of AI infrastructure. The panel explored what it takes to make inference faster, more efficient, and economically viable at scale, where the biggest opportunities are emerging, and what builders and investors are watching as AI workloads continue to grow.The discussion covered the infrastructure powering the inference boom, the economics behind large-scale AI deployment, and the opportunities developing across the inference stack.Panelists:Debajyoti (Debo) Datta, Co-Founder, Hippocratic AILance Matthews, Partner, DTCP GrowthAaron Jacobson, Partner, NEAAlberto Taiuti, CEO, ReactorWhat changes as AI inference moves from model development to massive-scale deploymentThe infrastructure, efficiency, and economics driving the next phase of AIWhere investors see the biggest opportunities across the AI inference stackModerator: Grace Gong, Founder & CEO, Smart Venture MediaTopics discussed:Smart AI 100 brings together founders, C-level executives, investors, and technology leaders shaping the future of AI.#AI #ArtificialIntelligence #Inference #AIInfrastructure #SmartAI100

Ori Goshen is the Co-founder and CEO of AI21 Labs, the Israeli AI company founded in 2017. AI21 has built its own foundation models and AI products used by millions, including Wordtune, and today it offers enterprises Lab-as-a-Service: their own frontier AI lab.~~~~~~~~~~~~~This episode is brought to you by Nebius — the ultimate cloud for AI innovators.Nebius provides AI infrastructure you can count on, combining reliability and speed with flexibility and engineering support unmatched by hyperscalers.AI leaders like Meta, Shopify, and Higgsfield already partner with Nebius to run their AI workloads. Plus, venture-backed startups can save up to $150,000 on compute costs when they apply for access. Visit nebius.com or nebius.com/startups to learn more~~~~~~~~~~~~~AI21's forward-deployed researchers, most of them PhDs and MSc holders, work embedded with a customer's team to train a specialized model and its agent harness together on the company's proprietary data. It's all built on AI21's platform for evaluating, training and serving models: eval suites, data generation pipelines, post-training infrastructure for supervised fine-tuning and reinforcement learning against task-specific rewards, harness optimization, and optimized inference. Customers own the resulting weights and harness, along with a documented, reproducible training recipe, and projects take weeks rather than quarters. Before AI21, Ori co-founded Crowdx, a network analytics company acquired by Cellwize, which Qualcomm later acquired.TopicsWhy enterprises want their own AI lab, and what that takesPost-training small, specialized models that outperform general frontier models on real tasksOptimizing the harness around a model, not just the model itselfWhy the next wave of AI gains will come from AI improving the systems that run AI#ArtificialIntelligence #AI21Labs #AIAgents #LargeLanguageModels #EnterpriseAI

AI is creating a new business model: AI-native services companies that use technology to make traditionally difficult-to-scale professional services faster, more efficient, and more accessible.In this fireside chat, Dan Mishin, Founder of Manifest, and Jake Saper, General Partner at Emergence Capital, explore why AI-native services are becoming venture-backable and why the services layer could represent the next major frontier for venture capital.The conversation examines what distinguishes AI-native services companies from traditional software businesses and AI rollups, and why building from scratch can offer an advantage over retrofitting legacy businesses. Dan shares how Manifest is using AI to transform immigration law, while Jake explains what makes the model compelling from an investor's perspective.They also discuss how AI changes the economics of services businesses, the opportunities across industries such as legal and accounting, expanding access to professional services, and the importance of human adoption in scaling these companies.As software becomes easier to build, could the next major venture opportunity be in transforming the services themselves?Panelists: Dan Mishin, Founder, Manifest; Jake Saper, General Partner, Emergence Capital#AI #ArtificialIntelligence #VentureCapital #AIInvesting #FutureOfWork

Austin Spiegel is the co-founder of Sift, where he leads engineering, technical strategy, and go-to-market. ~~~~~~~~~~~~~This episode is brought to you by Nebius — the ultimate cloud for AI innovators.Nebius provides AI infrastructure you can count on, combining reliability and speed with flexibility and engineering support unmatched by hyperscalers.AI leaders like Meta, Shopify, and Higgsfield already partner with Nebius to run their AI workloads. Plus, venture-backed startups can save up to $150,000 on compute costs when they apply for access. Visit nebius.com or nebius.com/startups to learn more~~~~~~~~~~~~~He studied computer science at the University of Southern California before joining SpaceX, where he spent nearly five years building mission-critical software across rocket and spacecraft manufacturing, and tools for developing and operating the Starlink satellite constellation. After a year as a software engineering manager at Riot Games, Austin reconnected with co-founder Karthik Gollapudi and began building Sift from the conviction that the data infrastructure behind SpaceX's engineering culture should be available to every hardware company. He is based in Los Angeles.TopicsApplying lessons from SpaceX's engineering and telemetry systems to help hardware companies understand complex, real-world systemsBuilding data infrastructure for aerospace, robotics, and hardware teamsScaling a deep-tech startup by translating internal engineering tools into a better platform#Sift #Aerospace #DataInfrastructure #HardwareEngineering #DeepTech

Jerry Tworek is the Co-Founder and CEO of Core Automation, an AI research lab focused on automating AI research and developing new approaches to machine learning. ~~~~~~~~~~~~~This episode is brought to you by Nebius — the ultimate cloud for AI innovators.Nebius provides AI infrastructure you can count on, combining reliability and speed with flexibility and engineering support unmatched by hyperscalers.AI leaders like Meta, Shopify, and Higgsfield already partner with Nebius to run their AI workloads. Plus, venture-backed startups can save up to $150,000 on compute costs when they apply for access. Visit nebius.com or nebius.com/startups to learn more~~~~~~~~~~~~Before founding Core Automation in 2026, Jerry spent nearly seven years at OpenAI, where he rose to Vice President of Research. His work spanned neural program synthesis, reinforcement learning, and language model reasoning, with major contributions to ChatGPT, GPT-4, Codex, and the development of OpenAI’s o1 and o3 reasoning models. At Core Automation, he is pursuing a new vision for AI development: building systems that automate research workflows while exploring learning algorithms and architectures beyond conventional large-scale pretraining. His work focuses on how AI systems can become better at reasoning, learn from experience, and help accelerate the research process itself.TopicsWhat it takes to build systems that can solve increasingly complex problems.Automating AI research with agents, including how systems work.Moving beyond conventional pretraining toward continual learning.#ArtificialIntelligence #AIResearch #AgenticAI #MachineLearning #AIReasoning

Vitaly Gordon is the Co-founder and CEO of Faros, a company working at the intersection of AI and software engineering to help organizations manage agentic software factories. ~~~~~~~~~~~~~This episode is brought to you by Nebius — the ultimate cloud for AI innovators.Nebius provides AI infrastructure you can count on, combining reliability and speed with flexibility and engineering support unmatched by hyperscalers.AI leaders like Meta, Shopify, and Higgsfield already partner with Nebius to run their AI workloads. Plus, venture-backed startups can save up to $150,000 on compute costs when they apply for access. Visit nebius.com or nebius.com/startups to learn more~~~~~~~~~~~~~Before founding Faros, Vitaly served as VP of Engineering at Salesforce, where he founded Salesforce Einstein, one of the first comprehensive enterprise AI platforms. His experience spans engineering leadership, enterprise AI, and building systems that bring machine learning into real-world business workflows. At Faros, he is focused on helping organizations understand and manage the growing role of AI agents in software development, as engineering teams move toward increasingly automated workflows. As AI agents take on more coding and engineering tasks, companies face new challenges in measuring productivity, coordinating work, and maintaining visibility across increasingly complex software delivery processes.TopicsManaging agentic software factories and measuring AI's impact on engineering productivityPositioning and scaling AI infrastructure startups through effective go-to-market strategiesHow AI engineering intelligence connects developer workflows, productivity metrics, and business outcomes#ArtificialIntelligence #AIAgents #SoftwareEngineering #DeveloperProductivity #StartupGrowth

Dr. Juha Vartiainen is the Co-Founder and Chief Global Affairs Officer of IQM Quantum Computers, a European quantum computing company developing full-stack superconducting quantum computers. ~~~~~~~~~~~~~This episode is brought to you by Nebius — the ultimate cloud for AI innovators.Nebius provides AI infrastructure you can count on, combining reliability and speed with flexibility and engineering support unmatched by hyperscalers.AI leaders like Meta, Shopify, and Higgsfield already partner with Nebius to run their AI workloads. Plus, venture-backed startups can save up to $150,000 on compute costs when they apply for access. Visit nebius.com or nebius.com/startups to learn more~~~~~~~~~~~~He co-founded IQM in 2018 as a spin-out from Aalto University and VTT Technical Research Centre of Finland, serving as Chief Operating Officer before moving into global affairs. In his current role, Juha works at the intersection of quantum technology, industrial strategy, and geopolitics, engaging with governments, policymakers, and strategic partners to advance quantum computing adoption and strengthen sovereign quantum capabilities. With a PhD in quantum computing and a background spanning software engineering, medical diagnostics, and space technology, he brings both technical and operational experience to the challenge of commercializing advanced computing. His work focuses on translating scientific breakthroughs into scalable quantum systems and helping shape Europe’s role in the emerging global quantum economy.TopicsMoving quantum computing from scientific research to commercially deployable systemsBuilding a global quantum industry through government partnerships, sovereign capabilities, and international technology strategyScaling deep-tech companies from university spin-outs to global businesses #QuantumComputing #DeepTech #QuantumTechnology #EuropeanTech #FutureOfComputing