Pre-Seed and Seed · AI Infrastructure · Physical AI · Deep Tech

434 VC is an early-stage fund backing technical founders building the next-generation of critical technologies in AI.

We invest at pre-seed and seed stage in the software that takes AI into production (AI infrastructure and deployment), the intelligence that models and runs the physical world (physical AI), and the software for new hardware and science underneath both (deep tech).

Thesis

The era of AI deploymentand physical intelligence.

AI access is everywhere as foundation models continue to commoditize, yet, full-scale enterprise adoption is lagging behind. The capital funding foundation models has moved to later-stage rounds, leaving the infrastructure layers below underfunded at pre-seed and seed. We invest in software infrastructure companies bridging that access-to-adoption gap with deployment, execution and reliability platforms that turn models into systems enterprises can trust and deploy at full scale.

Scaling foundation models into the physical world has made more obvious the need for latent systems intelligence that more closely mimics how physical systems function. Better simulation and prediction technologies result in higher asset utilization which means more output per unit of capital and material. We invest in companies that capture proprietary data from physical systems to predict, simulate and control them unlocking higher productivity gains.

The future of the physical economy depends on compute. AI compute is now constrained by energy, heat, memory bandwidth and fab capacity. Every era of computing was unlocked by a new material substrate such as silicon. We look to invest in software that discovers and designs the materials, chips and architectures on which the next generation of compute will depend on.

The three investment verticals we tackle create a reinforcing improvement loop where AI deployment turns models into working systems, physical intelligence uses them to run assets better and design new materials, and those materials unlock the compute that makes the next generation of intelligence possible.

01

AI infrastructure

Deployment, execution and reliability platforms that turn models into systems enterprises can trust.

02

Physical AI

Proprietary data from physical systems turned into models that predict, simulate and control them.

03

Deep tech

Software that discovers and designs the materials, chips and architectures for the next generation of compute.

Why 434 VC

We've built AIwhere mistakes matter.

We've spent the last decade building production AI in regulated, high-stakes industries, finance and healthcare especially, where trust, compliance, and accuracy aren't optional. We evaluate founders as builders who can read your architecture and product and tell you what's real, not spectators.

Goodfin (YC)Google (Search/Cloud/ML)Sprinter HealthAmazonAkamaiIntel
How We Invest

Technical founders in AI infrastructure, physical AI and deep tech.

We invest in West Coast and East Coast-based founders at pre-seed and seed stage with checks ranging from $100k to $500k. We follow-on with conviction and work closely with the companies we back.

AI infrastructurePhysical AIDeep techAI-native softwareTechnical foundersWest CoastEast Coast
What We Look For

What makes a companyworth a conversation.

01

Technical founders

Builders that demonstrate technical and scientific knowledge, deep domain expertise, strong commercial relationships in target markets, polymaths and original thinkers.

02

Commercial traction

Companies that demonstrate early product-market fit through completed or in progress pilots. We like to invest as those pilots turn into recurring contracts.

03

Sizable and fast-growing markets

Sizable and fast-growing markets to support a platform vision that compounds well beyond its first product.

04

Deep technology

Products with real technical depth, validated by experts and benchmarked against existing alternatives to prove outperformance.

05

Moats that compound

Infrastructure embedded in customer workflows and proprietary physical data generated inside factories, labs and grids create advantages that compound with every deployment.

06

Insider relationships

The team has insider relationships to win first target customers fast.

07

Measurable ROI

You show measurable ROI for your customers.

Active Themes

What we're tracking right now.

01

AI: Evaluation, Observability and Reliability

How AI behavior is measured, debugged, validated and controlled

02

AI: Data and Context Infrastructure

How AI accesses, retrieves and understands relevant information

03

AI: Agent Runtime and Developer Infrastructure

How AI systems execute actions and workflows safely and durably

04

Physical AI: World Models and Simulation

How a system's future is predicted and tested before anyone acts

05

Physical AI: Perception and Physical Data

How machines build an accurate picture of the physical world, and get the data to learn it

06

Deep tech: AI Compute Silicon and Architectures

How AI is computed at lower cost and energy per token

A point of view, not a prediction · Updated periodically

Physical AI

AI that powersphysical systems.

We invest in the intelligence and execution systems that understand and control the physical world.

Today's foundation models learned from text and pixels, a description of the world rather than the world itself. Most of what matters in the physical economy was never written down: how a turbine wears, how a cell responds to a compound, how a grid behaves under stress. We back founders building models that learn the hidden state of a physical system (a cell, a machine, a grid) and simulate how it will behave under different conditions.

That lets operators test decisions before committing capital, and lets the physical economy run on prediction instead of buffers: less idle capacity, less safety stock, less conservative maintenance. This data can't be scraped, it has to be generated. That makes the sensors, experiments and workflows that produce it the moat.

01

Perception and Physical Data

How machines build an accurate picture of the physical world, and get the data to learn it.

02

World Models and Simulation

How a system's future is predicted and tested before anyone acts.

03

Embodied Intelligence and Edge Execution

How predictions become safe action on the device, robot, vehicle or line.

04

Autonomous Discovery

How AI moves from predicting to finding new materials, molecules and biology.

434-backed · Case study

Ottonomy

Intelligence that has to work beyond the demo.

Ottonomy builds autonomous delivery robots for demanding indoor and outdoor environments. It is a clear expression of what we look for in physical AI: software, hardware, and operations designed as one system.

Visit Ottonomy
Why we invested

The hard part of autonomy is not making a machine move once. It is making the full system perceive, decide, and act reliably in environments that refuse to stay controlled.

At Google, we learned what separates an impressive model from a production machine-learning system: feedback loops, infrastructure, edge cases, and relentless attention to reliability. Physical AI raises that bar. Errors no longer live only on a screen; they meet people, objects, weather, latency, and changing terrain.

Nikhil Tyagi brings the complementary systems lens. Across Broadcom, Apple, Tesla, Meta, and Qualcomm, he has spent two decades taking connectivity from silicon to shipped products. That experience sharpens the question behind this investment: can every layer work together under real-world constraints, not just in isolation?

Ottonomy fits that conviction. Its robots bring perception, planning, control, edge intelligence, and fleet operations into one deployed system. Each real-world run can create the operational data needed to improve the next one. That is the kind of compounding loop we want to back.

Investment signals
Full-stack autonomy
Perception, planning, control, and operations built together.
Real deployment
A system designed for dynamic indoor and outdoor environments.
Edge reliability
Intelligence that must perform where latency and connectivity matter.
Data that compounds
Operational experience that can strengthen the deployed system over time.
What You Get

More than a check.

434 VC is built for founders who want a technical partner with conviction about the build, the workflow, and the domain — not a passive line on the cap table.

01

A Technical Partner Who's Shipped

We read your architecture, model choices, and product the way a builder who's shipped would. Honest reads, sharper questions.

02

Buyer and Design-Partner Relationships

We have relationships with buyers and operators across finance, wealth, and legal. Where we can open a door for a design partner or first customer, we will. Where we can't, we'll say so.

03

Founder Community and Deal Flow

Through House of 434, YC, and our operator network, you plug into the rooms where this work is already happening.

04

Invested Like a Builder

We've written checks and sat on the founder side. We know what helps early and what's noise.

How We Source

How we source.

Our deal flow is earned, not paid for. Through House of 434, the community we convene, plus the YC network as founders ourselves and a deep bench of ex-Google operators, we see strong technical founders early, often before they're raising.

Founder-Market Fit

The signal we weightmost heavily.

At the early stage, the single most important signal is founder-market fit: domain depth that goes beyond a deck, and conviction earned by living inside the workflow you're rebuilding.

We back founders who understand the buyer, the workflow, and the constraints of the industry better than anyone else trying to serve it.

Are We a Fit

A quick self-check.

Not every great company is a 434 VC company.

What gets us to pass
  • ✕AI wrapper without a moat
  • ✕No founder-market fit
  • ✕Unclear buyer
  • ✕A services business with no credible software path
  • ✕A team that can demo but can't ship to production
  • ✕No path to venture-scale outcomes
  • ✕Technology searching for a problem

These are not negotiable.

Why 434

434 began as an address. 434 Bayview was the home where many of the most important chapters of our story began. Looking back, what stood out was not the house. It was what emerged from it. Conversations that became friendships. Friendships that became collaborations. Collaborations that became companies.

434 VC is built on that belief. That meaningful things often start small. That the highest-signal ideas and the strongest relationships live in private rooms long before they become visible.

434 VC exists to back founders building the future of software through AI.

Team
Shilpi Nayak, Founder and Managing Partner of 434 VC
Founder and Managing Partner
Shilpi Nayak

Shilpi Nayak

Founder and Managing Partner, 434 VC

Shilpi Nayak is the Founder and Managing Partner of 434 VC.

She has spent the last decade building production AI in the industries where it's hardest to get right — finance, healthcare, and other regulated, high-stakes domains. As Co-Founder and CTO of Goodfin (YC), she built the AI platform from the ground up and co-led the launch of Goodfin Go, the system that opened access to the most sought-after private companies of this cycle — SpaceX, Anthropic, xAI, Cerebras — to a new generation of investors. Before that she spent four years at Google building large-scale machine learning across Search, Cloud, and user modeling, with earlier engineering work at Amazon, Akamai, and Intel.

Her work on AI in financial analysis has been featured in the Financial Times and CNBC. She has spoken at Money 20/20 and New York Tech Week, and has been an active angel investor for years, backing AI-native founders including Thomas.

She started 434 VC to back the operators she knows best: builders turning hard-won domain expertise into the next generation of AI-native companies.

House of 434

A community for founders, operators, and investors.

House of 434 is where founders, operators, and investors explore AI, capital, and company building together — intimate conversations, field notes, and gatherings around the questions shaping the applied-AI era.

House of 434 and 434 VC are separate: the community stands on its own, and we're the common thread behind both.

Visit House of 434
Contact

Building something we should know about?

We're always interested in meeting exceptional technical founders building ambitious companies across our areas of focus.

We review every submission and will reach out if there's a potential fit.

Submit Your Company