Working at Baseten in 2026: $5B Valuation, 100 People & the Inference Layer

There is a quiet, unglamorous layer of the AI stack that every foundation model, every chatbot, every image generator depends on: inference infrastructure. When a user sends a prompt to an LLM, something has to allocate the GPU, load the model weights, run the forward pass, stream the tokens back, and do it all in under a second at massive scale. That something, increasingly, is Baseten.

Founded in 2019 by Tuhin Srivastava, Baseten has become the infrastructure layer that AI companies build on. In January 2026, the company raised $300 million at a $5 billion valuation — a staggering number for a team of roughly 100 people. With $585 million in total funding from investors including IVP, CapitalG (Alphabet's investment arm), NVIDIA, Greylock Partners, and Spark Capital, Baseten is one of the most capital-efficient companies in the AI ecosystem. But what is it actually like to work there?

We pulled data from Baseten's company profile, employee reviews, and public sources to build the most detailed picture available of Baseten as an employer in 2026.

Baseten at a Glance

Those numbers tell a story on their own. A $5 billion valuation divided by ~100 employees works out to roughly $50 million per person — one of the highest ratios in all of tech, even though Baseten’s absolute valuation is dwarfed by giants like Anthropic ($965B post-money, ~5,000 employees) and Stripe (~$91B, ~8,000 employees). Baseten is in rarefied air on a per-employee basis, and it tells you something about both the leverage of infrastructure businesses and the intensity expected from each person on the team.

The Inference Thesis: Why Baseten Matters

To understand why Baseten is worth $5 billion, you need to understand the AI inference problem. Training a model gets the headlines. Inference — actually running that model in production, at scale, with low latency and high reliability — is where the real infrastructure challenge lives. Every time someone sends a message to ChatGPT, generates an image with Midjourney, or runs a voice call through an AI agent, inference infrastructure handles the heavy lifting.

Baseten's core product is an inference platform that handles GPU orchestration, model serving, autoscaling, and optimization. Their open-source framework, Truss, provides a standardized way to package and deploy ML models. Under the hood, they're solving problems that sit at the intersection of distributed systems, GPU programming, and low-level performance optimization — the kind of work that attracts engineers who care about the metal as much as the abstraction.

This is what makes Baseten's engineering challenges genuinely unique. You're not building another SaaS product. You're building the platform that other AI companies depend on to serve their models. When Baseten has a bad day, AI products across the industry feel it. That level of responsibility, combined with the technical depth required, is part of what draws engineers in — and part of what makes the work intense.

What Public Signals Tell Us About Culture

Baseten has very limited public review data — at the time of writing, only a single Glassdoor review exists (5.0 stars, from a software engineer). That makes traditional rating breakdowns unreliable for Baseten today. Instead, we leaned on the engineering blog, public podcast interviews with founders and engineers, social posts from the team, and employee discussions on Hacker News to triangulate what working there actually looks like.

What comes through consistently from those sources: a small, engineering-driven team operating at the frontier of ML inference, with high ownership and a fast-shipping culture — balanced against the intensity of a 100-person company growing into a market where every week brings new model architectures, new GPU constraints, and new customer demands. If you want a relaxed 9-to-5 environment, Baseten is probably not the right fit. If you want to work on hard problems alongside a small, intense team and can handle periods of sustained effort, read on.

What Makes Baseten's Culture Different

Most culture descriptions at this stage would talk about ping pong tables and unlimited PTO. Baseten's culture is defined by something more concrete: the compression of responsibility. At ~100 people with a $5 billion valuation, there is no room for passengers. Every engineer, every PM, every go-to-market person is carrying a disproportionate share of the company's output. This is what the many-hats value actually means in practice — not that you're understaffed (though resources are limited), but that the surface area of your impact is unusually large.

Three things stand out about Baseten's culture from employee reviews and public signals.

1. Engineering drives decisions

Baseten is an engineering-driven company in the truest sense. The product is infrastructure. The customers are engineers. The competitive moat is technical. This means that engineering decisions — what to optimize, what to build next, how to architect systems — are not downstream of product or sales decisions. They are the decisions. Engineers who have worked at sales-led companies will notice the difference immediately.

2. Flat hierarchy, direct access

With ~100 people, there are no layers. A flat structure isn't just a talking point — it's a mathematical reality when you have fewer employees than most companies have on a single floor. Engineers talk directly to the founder. Decisions happen in Slack threads and small meetings, not in multi-week approval chains. This is exhilarating for people who want agency and exhausting for people who want structure.

3. Ship fast, iterate in production

Baseten's culture is oriented around shipping quickly. In the AI inference space, the competitive landscape shifts monthly. New model architectures emerge, GPU availability fluctuates, customers need support for the latest open-source models within days of release. This creates a cadence where speed is a feature, not a compromise on quality. The team iterates rapidly, often working on optimizations that shave milliseconds off inference latency — because at scale, milliseconds are money.

What Public Signals Suggest About Day-to-Day Work

Synthesizing the engineering blog, public talks, podcast interviews and Hacker News discussions where Baseten engineers and the founder have participated, here are the themes that come up consistently.

What looks attractive

• Cutting-edge ML inference work. The technical surface area is genuinely novel: GPU kernel optimization, distributed model serving, autoscaling for spiky inference workloads, and supporting new open-source models within days of release.
• High ownership from day one. At ~100 people, individual contributors ship code that real customers depend on within weeks of starting. There is no large infrastructure team sitting between you and production.
• Direct product impact. Because Baseten's customers are themselves AI companies running models in production, optimizations show up as real, measurable latency reductions across customer deployments — a tight feedback loop is part of the appeal.
• Equity at a fast-rising private valuation. Baseten closed a $300M Series E at a $5B post-money valuation in January 2026, and press reports indicate the company is in talks for a follow-on round at roughly $11B. New-hire equity is expensive relative to early employees but still meaningful if the inference market continues to compound.

What looks hard

• Limited resources for the scope of the market. A ~100-person team selling into virtually every AI company means there is always more work than people to do it. Expect to wear multiple hats.
• Shifting priorities. The AI inference landscape changes monthly — new model architectures, new GPU availability dynamics, new customer needs. Roadmaps are subject to reprioritization.
• Process is still being built. Internal docs, engineering specs, and project management practices at this stage are often ad-hoc. Engineers who need well-defined frameworks can find this chaotic.
• Intensity. Infrastructure customers depend on uptime, and the company's pace is fast. This is not a 9-to-5 environment.

The pattern, if Baseten is anything like other high-growth infrastructure companies, is technical depth, ownership, and equity upside in exchange for intensity, ambiguity, and the inevitable growing pains of scaling faster than your processes. Similar themes appeared in early reviews of Vercel and Databricks at comparable stages.

Compensation & Equity

Baseten offers competitive base salaries for the AI infrastructure space, combined with equity that could be significant given the company's trajectory. At a $5 billion valuation with $585 million raised, employees are betting on continued growth — and the AI inference market gives them reason to. The total addressable market for inference is projected to dwarf training spend as AI adoption accelerates across industries.

The equity component is particularly interesting at Baseten's stage. Employees who join now are getting in at a $5 billion valuation — expensive relative to early employees, but potentially cheap if the company continues to grow with the AI inference market. For comparison, companies like Databricks ($62B valuation) went through a similar trajectory, and early-to-mid-stage employees saw significant returns. The risk profile is lower than a seed-stage startup but the upside is still meaningful — a rare combination.

Benefits are competitive for a company of this size: comprehensive healthcare, remote flexibility, and the intangible benefit of working on problems that are genuinely at the frontier of the AI infrastructure stack.

Engineering Culture & Tech Stack

Baseten's engineering culture is defined by the nature of the problem they're solving. ML inference infrastructure sits at the intersection of systems programming, GPU optimization, distributed computing, and machine learning — and the team needs to be fluent across all of these.

Tech Stack

The stack reflects the company's position in the infrastructure layer. Python and PyTorch for the ML-facing interfaces. Go for high-performance backend services. CUDA and Triton for GPU-level optimization. Kubernetes and Terraform for orchestration. This is not a typical web development stack — it's a systems engineering stack, and it attracts a specific type of engineer who cares about performance at the hardware level.

How engineering works at Baseten

• End-to-end ownership. Engineers own features from design through production. There's no "infrastructure team" that handles deployments for you — you build it, you ship it, you monitor it. With ~100 people total, there's no other option, and most engineers consider this a feature.
• Performance is a first-class concern. When your product is inference speed, every optimization matters. Engineers regularly work on GPU kernel optimization, batching strategies, and memory management. A 10% latency improvement at Baseten's scale affects thousands of deployments.
• Open source as a feedback loop. Truss, Baseten's open-source model serving framework, serves as both a product and a community signal. Engineers contribute to and maintain open-source projects, which keeps them connected to the broader ML engineering community.
• Rapid iteration with real customers. Because Baseten's customers are AI companies deploying models in production, the feedback loop is tight. You ship an optimization, customers see the impact immediately, and the metrics prove it. This direct product impact is one of the most cited reasons employees enjoy working there.

Baseten vs. Other AI Infrastructure Companies

Baseten operates in a competitive space. Here's how it compares to nearby companies in the AI infrastructure landscape.

For a detailed side-by-side with any company in our database, use the comparison tool.

Who Thrives at Baseten

Based on the culture signals, employee reviews, and the nature of the work, here's who tends to do well at Baseten — and who might want to look elsewhere.

• Systems engineers who care about performance. If you get excited about GPU memory hierarchies, kernel optimization, and shaving microseconds off inference latency, Baseten is one of the few places where that's the core product, not a side project. This is infrastructure work at the most demanding level.
• People who want ownership over structure. With ~100 people, there's no playbook. You'll define processes as much as follow them. If that sounds exciting, you'll love it. If it sounds stressful, you'll struggle. The flat hierarchy means your voice matters, but it also means nobody's going to tell you exactly what to do.
• Engineers who want to understand the full stack. At Baseten, wearing many hats isn't just encouraged — it's inevitable. You might be debugging a Kubernetes autoscaling issue in the morning and optimizing a CUDA kernel in the afternoon. If you want to specialize narrowly, a larger company is a better fit.
• People comfortable with ambiguity. The AI landscape shifts monthly. Baseten's priorities shift with it. If you need a clear 12-month roadmap, this pace will feel chaotic. If you're energized by adapting to new challenges, you'll thrive.
• Builders who want equity upside. At a $5 billion valuation with a market position in the fastest-growing segment of AI infrastructure, Baseten's equity has meaningful upside potential. Employees who joined at earlier stages have already seen significant paper gains. Joining now carries more risk than a public company but more upside than a late-stage unicorn.

Baseten is not ideal for people who prioritize work-life balance above all else (the 3.5 WLB score reflects real intensity), or those who want well-defined career ladders and structured mentorship programs. At ~100 people, those things are still being built. If you need them now, consider more established companies like HubSpot or Notion.

Open Positions at Baseten

Baseten currently has 71 open positions across engineering, infrastructure, and go-to-market roles. For a company of ~100 people, that represents significant planned growth — Baseten is looking to potentially double its team. Given the $300 million raised in January 2026, there's no shortage of runway to fund this expansion.

If the engineering challenges, ownership, and equity story described in this post resonate with you, now is one of the best times to join. The company is large enough to have product-market fit and real revenue, but small enough that every new hire meaningfully shapes the culture and the product.

For full details on Baseten's open roles, culture values, and side-by-side comparisons with other companies, visit the Baseten culture profile page.