Benny Chen is the cofounder of Fireworks AI, an AI infrastructure platform. They have raised $327M in funding from Benchmark, Sequoia, Lightspeed, Index, and others.

Benny's favorite book: Principles (Author: Ray Dalio)

(00:01) Intro and why AI infrastructure is having a moment
(00:06) Training vs inference: what’s working and where the real bottlenecks are
(01:25) Why inference is the hard problem in production
(03:30) What breaks at scale when AI systems hit real users
(05:29) GPUs, hardware constraints, and why power is now a first-class concern
(06:02) What you’re actually paying for in inference
(07:21) Reliability, compliance, and enterprise expectations
(09:49) Training and inference capacity: when they blur together
(11:06) How to make inference fast in practice
(13:06) System design choices behind modern inference platforms
(15:28) Inference economics and cost tradeoffs
(18:02) When fine-tuning actually makes sense
(21:58) What “best model” really means for real companies
(24:25) Production LLM architectures that actually work
(27:46) Building an AI infra company customers can trust
(29:27) Shipping fast without breaking reliability
(31:14) Go-to-market lessons for infra startups
(34:17) Where inference platforms are heading next
(36:32) Rapid fire round

--------
Where to find Benny Chen:

LinkedIn: https://www.linkedin.com/in/benny-yufei-chen-2238575a/

--------
Where to find Prateek Joshi:

Website: https://prateekj.com
Research Column: https://www.infrastartups.com
LinkedIn: https://www.linkedin.com/in/prateek-joshi-infinite
X: https://x.com/prateekj

Surojit Chatterjee is CEO of Ema, an agent platform build AI employees. They have raised $61M in funding from Accel, Section 32, and others. Before Ema, he was the chief product officer at Coinbase. And before that, a VP at Google.

Surojit's favorite book: Man's Search for Meaning (Author: Viktor Frankl)

(00:01) Welcome
(00:07) Defining the “AI Employee”
(02:23) Lessons from Google: Building for Scale
(06:59) Coinbase CPO: Hypergrowth & Product Leadership
(09:24) Market Framing: Why “AI Employee” vs Copilot
(14:29) Platform Building Blocks (Agents, Orchestrator, Fusion, Governance)
(19:26) Trust, Security, and On-Prem Deployment
(23:11) Model of Models: How Fusion Picks & Combines LLMs
(29:10) What Infra Is Still Missing (Eval at Scale, Speed)
(32:10) Rapid Fire Round

--------
Where to find Surojit Chatterjee:

LinkedIn: https://www.linkedin.com/in/surojitchatterjee/

--------
Where to find Prateek Joshi:

Website: https://prateekj.com
Research Column: https://www.infrastartups.com
LinkedIn: https://www.linkedin.com/in/prateek-joshi-infinite
X: https://x.com/prateekj

Rishi Bhargava is CEO of Descope, an identity management platform for customers and AI agents. They've raised $88M in funding from investors such as Notable Capital, Lightspeed, Unusual Ventures. The two previous he founded were acquired by Palo Alto Networks and McAfee.

(00:01) Introduction
(00:08) Origin story: why identity and passwords needed a rethink
(02:59) Passwords vs passkeys explained in plain English
(05:06) Why logging in is still painful (and why passwords persist)
(09:06) Account takeovers explained: how hacks actually happen
(11:59) Building security products: philosophy vs regular software
(14:24) The ideal login experience: from frustration to seamless access
(16:40) What is an AI agent? Defining agent identity simply
(21:54) Good bots vs bad bots: trust, access, and control in an agent world
(25:03) Breaches and blast radius: security before vs after Descope
(27:55) Company building lessons from Demisto to Descope
(30:15) AI trends that matter most for enterprise products
(32:40) Rapid Fire Round

--------
Where to find Rishi Bhargava:

LinkedIn: https://www.linkedin.com/in/bhargavarishi/

--------
Where to find Prateek Joshi:

Website: https://prateekj.com
Research Column: https://www.infrastartups.com
LinkedIn: https://www.linkedin.com/in/prateek-joshi-infinite
X: https://x.com/prateekj

Gou Rao is CEO of NeuBird, an agentic AI Site Reliability Engineer for IT teams. They've raised $44.5 Million from Mayfield and M12. He was previously the CTO of Citrix and Portworx.

(00:01) Introduction
(01:07) What Does an SRE Do?
(02:19) Inside a Typical Incident Flow
(04:16) What Can Be Automated?
(05:52) Deploying Hawkeye: Day 1 to Day 100
(11:59) Earning Trust for Autonomous Agents
(14:57) Versioning Agent Behavior & Chain of Thought
(17:02) Building Agentic Infra Products
(18:38) Access Control for Agents
(20:29) Company Building in the AI Era
(23:53) Competitive Edge in AI + Infra
(26:35) Model Choice & Agent Reasoning Quality
(29:33) Biggest Product Bet
(31:22) Exciting AI Advancements
(33:04) Rapid Fire Round

--------
Where to find Gou Rao:

LinkedIn: https://www.linkedin.com/in/gouthamrao/

--------
Where to find Prateek Joshi:

Research Column: https://www.infrastartups.com
Newsletter: https://prateekjoshi.substack.com
Website: https://prateekj.com
LinkedIn: https://www.linkedin.com/in/prateek-joshi-infinite
X: https://x.com/prateekj

Brian Moore is CEO of Voxel51, a data infra platform for visual AI. They most recently raised a $30M Series B led by Bessemer.

Brian's favorite books: Trillion Dollar Coach (Author: Eric Schmidt, Jonathan Rosenberg, and Alan Eagle)

(00:01) Introduction and setup
(00:22) Defining visual AI — beyond traditional computer vision
(02:14) Why visual data is so hard to manage
(04:17) Common “gotchas” in image and video datasets
(06:43) Is it a data problem or a model problem?
(09:41) The importance of edge cases and scenario analysis
(10:46) Coverage and handling rare events in datasets
(13:35) Using synthetic data and foundation models to fill data gaps
(14:25) The origin story of Voxel51 and the birth of FiftyOne
(17:56) Open source strategy and community growth
(19:31) Handling massive visual datasets — storage best practices
(22:03) Cost vs. quality tradeoffs in video storage
(23:54) Cleaning and indexing messy datasets
(25:49) Measuring real progress — beyond simple metrics
(27:40) Compute bottlenecks and faster iteration loops
(30:05) The economics of data infrastructure
(31:53) Labeling inefficiencies and smarter annotation workflows
(33:56) Hidden costs of data wrangling and wasted engineering time
(35:10) Positioning Voxel51 and lessons for founders
(37:53) The future of visual AI and missing industry standards
(40:36) Rapid Fire Round

--------
Where to find Brian Moore:

LinkedIn: https://www.linkedin.com/in/brimoor/

--------
Where to find Prateek Joshi:

Research Column: https://www.infrastartups.com
Newsletter: https://prateekjoshi.substack.com
Website: https://prateekj.com
LinkedIn: https://www.linkedin.com/in/prateek-joshi-infinite
X: https://x.com/prateekvjoshi

Carina Hong is CEO of Axiom Math, where they're building a self-improving superintelligent reasoner, starting with an AI mathematician. She's a Rhodes Scholar, first-gen college grad and mathematics prodigy who earned dual degrees in mathematics and physics from MIT in 3 years. And a joint JD/PhD at Stanford. They just raised a $64M seed round from B Capital, Greycroft, Madrona, and Menlo Ventures.

Carina's favorite books: Proofs from THE BOOK (Author: Martin Aigner, Günter M. Ziegler)

(00:02) Intro
(00:38) What self-improving mathematical superintelligence means
(04:04) Proofs as programs: Lean and the data gap
(06:36) How AI proves: human-style vs. Lean-style reasoning
(10:43) Carina’s journey: from Olympiad problem-solver to theory-builder
(14:47) The engine room: data, infra, and building a math knowledge graph
(17:42) Verifying results: compile checks vs. LLM judges
(18:56) Self-improvement loops: skills libraries, memory, and conjecture↔prover curricula
(21:30) Synthetic data & auto-formalization strategy
(24:00) Benchmarks that matter: miniF2F, CombiBench, miniCTX v2
(26:24) Why combinatorics is uniquely hard for AI
(31:13) Compute footprint & scaling philosophy
(32:20) In-house Lean tooling and productization path
(33:57) Early use cases: formal verification in hardware/software
(36:19) Team blueprint: AI, programming languages, and math
(37:35) Scaling laws, efficiency, and bottlenecks
(38:26) If Axiom works: what becomes cheaper/faster for the world
(40:22) Rapid Fire Round

--------
Where to find Carina Hong:

LinkedIn: https://www.linkedin.com/in/carina-hong/

--------
Where to find Prateek Joshi:

Research column: https://www.infrastartups.com
Newsletter: https://prateekjoshi.substack.com
Website: https://prateekj.com
LinkedIn: https://www.linkedin.com/in/prateek-joshi-infinite
X: https://x.com/prateekvjoshi

Mukund Jha is CEO of Emergent, an agentic vibe-coding platform. They've raised $23M from Lightspeed, Y Combinator, Together Fund, and Prosus. He was previously the cofounder and CTO of Dunzo, a hugely popular ecommerce company in India.

Mukund's favorite books: The Hard Thing About Hard Things (Author: Ben Horowitz)

(00:01) Intro
(00:07) State of vibe-coding and where we are today
(01:42) Emergent in plain English: what the product delivers
(03:07) From prototype to traction: the first 90 days
(06:03) What changed in the last 24 months (models + infra)
(08:13) Early infra bets that enabled speed
(12:07) Precision vs. control: editing and debugging without code
(14:21) One-click to production: the unglamorous infra behind it
(15:55) Points of failure across prompt → plan → code → test → deploy
(17:53) Models division of labor: planning, codegen, tests, commits
(20:05) What “reasoning” means and how they evaluate it
(22:13) Context & memory strategy (beyond naive RAG)
(24:22) Representing large codebases so agents don’t hallucinate structure
(27:03) Orchestration walkthrough: adding SSO end-to-end
(29:40) Agent coordination protocols (how agents talk)
(31:05) Debugging long-running agents and trace observability
(32:37) Company-building lessons from Dunzo to Emergent
(36:10) Philosophy: offloading decisions to models
(36:57) Rapid Fire Round

--------
Where to find Mukund Jha:

LinkedIn: https://www.linkedin.com/in/mukund-jha-a1596413/

--------
Where to find Prateek Joshi:

Newsletter: https://prateekjoshi.substack.com
Website: https://prateekj.com
LinkedIn: https://www.linkedin.com/in/prateek-joshi-infinite
X: https://x.com/prateekvjoshi

Stefano Ermon is the cofounder of Inception Labs and an associate professor at Stanford. Inception is developing a new type of AI models called Diffusion LLMs.

Stefano's favorite book: If on a Winter's Night a Traveler (Author: Italo Calvino)

(00:01) Introduction
(00:38) What are autoregressive LLMs and how do they work
(02:28) How diffusion LLMs rethink generation
(04:02) The ceiling of autoregressive LLMs: cost, latency, reliability
(06:19) Why diffusion LLMs are commercially viable now
(09:12) Parallel refinement: how diffusion models generate text
(12:05) Understanding diffusion steps and efficiency
(13:49) Hardest engineering challenges at Inception
(15:23) From research to production: the power of data
(16:24) Where diffusion LLMs still lag behind
(18:18) Evaluations and benchmarks for diffusion LLMs
(20:20) Developer experience and OpenAI-compatible API
(21:47) Economics and GPU efficiency
(23:38) Hardware and runtime stack
(24:58) Competition and the evolving diffusion LLM landscape
(27:01) Where diffusion will win first — coding and agentic systems
(30:13) How diffusion changes infra, serving, and hardware design
(33:04) What’s next at Inception: reasoning and multimodality
(35:20) Rapid Fire Round

--------
Where to find Stefano Ermon:

LinkedIn: https://www.linkedin.com/in/ermon/

--------
Where to find Prateek Joshi:

Research column: https://www.infrastartups.com
Newsletter: https://prateekjoshi.substack.com
Website: https://prateekj.com
LinkedIn: https://www.linkedin.com/in/prateek-joshi-infinite
X: https://x.com/prateekvjoshi