San Francisco attracted over 60% of global AI funding in 2025. The companies defining what the next decade of technology looks like — the model labs, infrastructure providers, and AI-native products — are concentrated in Hayes Valley and SoMa. This guide covers the verticals worth targeting, what makes SF startup jobs genuinely different, and how Underdog.io gets you in front of the right companies before the opening is ever posted.
The SF startup job market
The companies concentrated in Hayes Valley and SoMa — now literally called Cerebral Valley — aren't just startups competing in existing markets. They're building the infrastructure, models, and toolchains that every other startup in every other city will run on for the next decade. An engineer working on AI training infrastructure, model evaluation, or inference optimization at an SF startup in 2026 is doing work that has no equivalent anywhere else. The problems are new. The playbooks don't exist yet. The scale is real.
The verticals worth targeting in 2026
SF's startup ecosystem in 2026 is more concentrated around AI than any city has ever been around any single technology shift. Here's how the verticals break down — and what each one means for engineering careers.
OpenAI, Anthropic, xAI, Safe Superintelligence, Mistral — the companies training the models everything else runs on. Engineering roles here include pre-training infrastructure, RLHF systems, evaluation frameworks, safety tooling, and inference optimization. These are the hardest engineering problems in the industry, with compensation to match. Most are hiring selectively but consistently.
Baseten, Anyscale, Scale AI, CoreWeave, Cerebras, Anysphere (Cursor) — the picks and shovels of the AI stack. Data infrastructure, model serving platforms, evaluation tooling, observability, and developer environments. These companies are growing because every AI company needs them. Engineers here work at the layer just below the model — where the reliability, speed, and economics of AI deployment are actually determined.
ElevenLabs, Replit, Perplexity, Runway, Character.ai, Hebbia — AI products with real users, real revenue, and real frontend and product engineering challenges. The React engineer who understands streaming UIs, the Python engineer who owns the API layer that calls five models — these roles exist at scale in SF in a way they don't elsewhere. The product problems are genuinely new.
Waymo, Bedrock Robotics, Nuro, Physical Intelligence — embodied AI is the next major wave after software AI, and SF has a density of robotics and hardware startups that no other city can match. Engineers working at the intersection of ML, systems programming, and physical systems are extremely rare and extremely well compensated. Not the most accessible entry point, but the highest ceiling.
SF startup salary guide 2026
SF leads the US in base salary for engineers. Every Underdog introduction includes the salary range — no mystery compensation after three rounds.
Sources: Glassdoor SF, ZipRecruiter SF, Wellfound SF Bay Area, Levels.fyi 2025.
Base salary at SF/Bay Area venture-backed startups. Equity is additive — at well-funded AI infrastructure companies, equity can be worth multiples of the cash comp over a 4-year vest. All Underdog introductions include the actual salary range in the first message.
The access problem
In SF's AI startup market, the roles that matter go fast — through networks, warm intros from YC partners, and curated platforms that founders trust. A job board posting for an AI infrastructure engineer role at a top SF startup will get 600 applications in 48 hours. That's not where the hire comes from. The hire comes from an engineer who was already visible to the team before the search started.
How Underdog gives you access
You tell us your stack and what you're looking for. We match you to SF startups whose open roles fit your background. You receive introductions — you don't apply to anything publicly.
Every SF startup that reaches out includes the salary range in the first message. No mystery comp discovery after a technical screen and a system design round.
Your profile is never public. Your current employer is blocked automatically. You can toggle off any specific company. SF's startup engineering community is small — discretion matters.
Most accepted engineers get at least one direct SF startup intro within 7 days. You decide which conversations go further. You control who gets your time.
SF and Bay Area startups in the network:
Seed through Series B. Not all are hiring at all times — we match you to what's active when you join.
Every time I use Underdog.io I remember that job searching doesn't have to be terrible. Thanks for the product.
Zach B. — Senior Software Engineer
Common questions
The private network for SF's best startup jobs
SF's top AI, infrastructure, and product startups hire engineers through Underdog. One profile. Salary first. Your employer blocked automatically. Free for engineers.
Get access to SF startup roles →