9,000 ML Jobs Are Open in SF. Why Can’t Engineers Get Hired?

If you walk through SoMa on any given Tuesday morning, you’ll pass at least two new AI startups that didn’t exist six months ago. The sidewalks around Mission Bay are busier than they’ve been since 2021. Cranes are back. Lease rates are climbing. By every surface-level metric, San Francisco is booming again.

But if you’re a software engineer who has spent the last three months sending applications into the void, that picture probably doesn’t match your experience. The city’s tech sector actually lost around 4,500 jobs in the information sector last year, even as AI investment in the Bay Area spiked. That tension between visible prosperity and real difficulty landing a role is the defining feature of the SF engineering market right now, and understanding how it works is the first step toward navigating it.

The market split into two lanes, and the dividing line keeps moving

There used to be a reasonably predictable career ladder in San Francisco tech. You started at a mid-size company or an early-stage startup, built your skills, and moved up or across as your interests evolved. Generalist engineers were valued because teams needed people who could wear multiple hats. That dynamic rewarded curiosity and breadth.

The current landscape looks different. AI-focused roles are multiplying so fast that the Bay Area now has more than 9,000 open machine learning engineering positions, with median total compensation sitting above $240,000. Senior roles at well-funded AI companies regularly clear $500,000 when you factor in equity, and a handful of positions at places like Anthropic or Stripe are reaching north of $800,000. If you have production experience building, fine-tuning, or deploying large language models, inference pipelines, or agentic systems, the market is practically fighting over you.

Meanwhile, the traditional software engineering track has contracted. Teams are smaller. Hiring cycles are longer. The growth-at-all-costs philosophy that defined the 2020 era has been replaced by what one industry analyst called “precision hiring,” where companies want fewer engineers who can each do more. AI-assisted development tools have accelerated this shift, because a team of ten with strong AI tooling can now produce output that used to require twenty.

That doesn’t mean non-AI engineering jobs are disappearing entirely. Companies still need people to build and maintain infrastructure, ship frontend features, improve developer tooling, and handle the thousand other things that keep products running. But the volume of those roles has shrunk, and the competition for each one has increased. If you’re applying for a standard full-stack position at a Series B startup, you’re competing against a much deeper applicant pool than you would have been a couple of years ago.

The geography still matters, but not the way it used to

One of the more interesting shifts in SF hiring is how the physical footprint of the tech industry has rearranged itself. Hayes Valley, sometimes referred to as “Cerebral Valley” by the AI crowd, has become the unofficial living room of the frontier AI community. The density of AI founders, researchers, and engineers in that neighborhood creates a kind of ambient networking that doesn’t really exist anywhere else. If you grab coffee at Sightglass on a weekday afternoon, there’s a reasonable chance the person at the next table is working on something related to multimodal models or reasoning architectures.

The Peninsula still houses the established players. Palo Alto and Mountain View remain home to Google DeepMind, Tesla’s AI division, and a constellation of enterprise AI companies. South Bay has its own cluster of semiconductor and robotics firms. But the energy, particularly for early-stage startups, has decisively shifted back into the city.

What’s changed is that geography alone doesn’t open doors the way it once did. During the last boom, simply being present in SF and attending a few meetups could generate warm introductions that led to interviews. Now, the sheer number of engineers concentrated in the city means that physical proximity is table stakes. The signal-to-noise ratio for both candidates and companies has gotten worse, which makes structured ways of connecting more valuable than casual ones.

What the companies hiring actually want has changed

If you look at job listings from SF-based AI startups, you’ll notice a pattern in what they’re asking for that differs from the requirements of a few years back. Python still dominates for ML-related work, but the surrounding expectations have expanded. Companies want engineers who understand retrieval-augmented generation, who have opinions about evaluation frameworks for LLM outputs, who can reason about prompt engineering as a systems design problem rather than a gimmick.

On the infrastructure side, containerized environments, CI/CD pipelines built for rapid model iteration, and experience with MCP (model context protocol) servers are showing up in listings with increasing frequency. The tooling ecosystem around AI development is evolving so quickly that familiarity with specific frameworks matters less than demonstrating you can learn and adapt to new ones.

The startups that are hiring most aggressively also care deeply about your ability to scope ambiguity. Early-stage AI companies often don’t know exactly what they’re building yet. They need engineers who can take a loosely defined problem, break it into testable hypotheses, ship something small, and iterate based on what they learn. That skill has always been valuable at startups, but in the current environment, where the underlying technology is changing month to month, it has become essential.

For engineers who aren’t specifically working in AI or ML, the market is asking you to demonstrate a different kind of value: either deep specialization in a domain that AI tools can’t easily replicate (complex distributed systems, security, database internals, performance engineering), or strong evidence that you can use AI-assisted tools effectively to multiply your output. The engineers who are struggling most in the current market are the ones positioned in the middle, competent generalists who haven’t yet differentiated in either direction.

The actual path forward for engineers in this market

If you’re reading this from a San Francisco apartment, trying to figure out your next move, the honest answer is that the market rewards intentionality more than it rewards persistence right now. Sending out a hundred applications and hoping for callbacks is one of the least efficient strategies available to you. The companies hiring are drowning in inbound applications and most of them lack the capacity to thoughtfully evaluate every one.

What works better is finding ways to get in front of hiring teams through channels where you’ve already been filtered for quality. Curated talent marketplaces, strong referral networks, and communities where engineers and startups interact directly have all become more important than they were a few years back. When a company knows that every candidate they’re meeting has already been vetted for technical skill and cultural fit, the conversation starts at a much higher level, and it moves faster.

Building in public also helps more than it used to. Contributing to open source projects in the AI tooling space, writing about technical problems you’ve solved, or shipping side projects that demonstrate your ability to work with new technologies all create the kind of signal that hiring managers notice. SF startups tend to value evidence of what you’ve built over credentials on your resume.

If you want to explore AI startup roles specifically, the Y Combinator ecosystem is a useful lens. There are currently more than 350 YC-backed AI companies in the Bay Area alone, and many of them are actively hiring across engineering, infrastructure, and product engineering disciplines. These aren’t all research labs. Plenty of them need full-stack engineers, backend engineers, and devops specialists who can build reliable products around AI capabilities.

The bottom line

San Francisco’s engineering market is legitimately strange right now. The city is the undisputed global center of the AI boom, and that boom is producing enormous demand for a specific slice of the engineering population. For everyone else, the market is harder than it has been in years, but not because there’s no opportunity. The opportunity is there. The challenge is finding the roles that match your skills and getting in front of the right companies before they’ve filled the position through someone’s Slack DM.

The engineers who are navigating this well tend to share a few traits: they’re specific about what they want, they invest time in communities where real hiring conversations happen, and they’re thoughtful about how they present their work to a market that has become much more selective.

If you’re an engineer exploring what’s out there in SF’s startup ecosystem, Fonzi connects vetted software engineers with VC-backed AI startups and tech companies through a curated matching process. It’s worth checking out the open roles and seeing what aligns with where you’re headed.