Cap Table Decoded
Cap Table DecodedHow Stage-Aware Talent Networks Help ML Engineers Vet Equity Terms Before the Offer Stage
Equity NormsLong read

How Stage-Aware Talent Networks Help ML Engineers Vet Equity Terms Before the Offer Stage

Stage, role and equity context let ML engineers judge offers early.

· 9 min read

ML engineers looking at Series A and Series B AI startup jobs need to know the funding-stage context, role benchmarks, and equity terms before the offer arrives. A headline ownership percentage cannot tell them whether a package is fair. Stage-aware talent networks help by surfacing that context before the first screen, when engineers can still use it to judge the opportunity.

Why ML engineers keep getting blindsided by equity at the offer stage

The offer arrives, and the engineer finally sees an equity percentage but has little to go on when judging it. Interviews have established technical fit, but not whether the grant fits the company’s funding stage, hiring plan, or actual role. Now the candidate must negotiate while building that frame of reference. The surprise is structural. A quick search can explain stock options, but it cannot reveal the company-specific assumptions behind the grant.

AI labs, large tech companies, and early-stage startups hire from the same talent pool, but their equity comes with different risks and ways to turn it into money. Experience with one system does not automatically transfer. As AI-focused startups bring more candidates into the founding-engineer funnel, more people face packages they have little practice judging.

What should an ML engineer look for in a Series A versus Series B startup job?

A headline percentage is a fraction. Its meaning depends on the company behind it and the position being filled. For Series A and Series B opportunities, the useful comparison starts with funding history, headcount, role scope, and the equity range for that role. The round label alone is not enough. Two companies at different stages should not be judged against one undifferentiated startup benchmark.

A Series A grant therefore cannot be ranked against a Series B grant just by choosing the larger percentage. The engineer needs to know how each position fits the company’s hiring plan and which role category its budget reflects. Both stages typically offer options, so moving from Series A to Series B does not by itself remove the exercise decision. Later-stage grants, typically at Series C+, are more often RSUs, a different compensation structure.

CRV’s founding engineer guide puts the median first-engineer grant at roughly 1.5% on a four-year schedule and notes that grants drop quickly as headcount grows. That benchmark belongs to a particular hiring position. It is not a standing entitlement for anyone with a founding-engineer title, nor a general benchmark for Series A or Series B ML hires. Looking up a percentage after receiving an offer helps only if the candidate already knows the stage, role, and ownership context needed to interpret it.

The equity mechanics that change actual value underneath the headline number

Even the right benchmark is only a starting point. Options require the engineer to pay the strike price to own shares, introducing an upfront cash decision and tax complexity. RSUs vest as ordinary income, and the engineer has no exercise decision to make. These instruments have different risk and upside profiles. A percentage that leaves the instrument unspecified leaves a central part of the bargain unresolved.

Dilution then changes the ownership position over time. An early-stage grant may represent significantly less ownership by exit, and a planned option-pool expansion can change the denominator before the engineer has settled into the role. Two offers with the same percentage may thus leave the engineer with different stakes in the future. The engineer needs to know what each percentage includes. Refresh grants also change the stake, so a refresh policy in writing tells the engineer more than a vague promise of future equity.

The exercise window determines how quickly an employee must act after leaving. A 90-day post-termination window is a common default, while some companies extend it to one to five years. That difference can force or postpone a costly decision. Tax exposure can make the decision harder still. An engineer who exercises ISOs without first working out the Alternative Minimum Tax exposure can face a large cash tax bill on paper gains with no secondary window open.

Some senior hires negotiate milestone triggers tied to model deployment thresholds, inference cost targets, or revenue attribution. Those terms can compress the value-realization timeline. None of these mechanics is visible in an ownership percentage alone. Learning about them only after the offer creates so much work at the wrong moment.

How role title confusion compounds equity mispricing for ML engineers specifically

Before choosing a benchmark, the engineer must know what the job involves. An AI engineer typically turns existing models into a useful product, an ML engineer makes machine-learning systems reliable and scalable in production, and a research scientist creates or validates new methods. The boundaries overlap, but the compensation bands differ. A title that understates the work can place the candidate in the wrong budget before compensation is ever discussed.

That mismatch has a direct financial consequence. Advertised base pay for applied AI engineering sits materially below frontier research bands, so a company seeking model-training work under an “AI engineer” title may budget against the wrong market. LLM engineering adds another unsettled category, with compensation still sorting itself out. A familiar title is not a reliable benchmark. The comparison must fit the work the engineer will do.

Most Series A and Series B AI startups need a research engineer. Meanwhile, ML platform work involving Kubernetes, serving infrastructure, and evaluation tooling belongs to a distinct hiring category whose scarcity can create pricing power. The engineer loses that advantage by accepting a benchmark for a different kind of work. The role category should be confirmed before the first technical screen. Stage-awareness and role-awareness work together, because neither alone establishes whether the equity range fits the job.

What a founding engineer gives up in salary — and why that makes equity vetting non-optional

CRV’s guide says founding-engineer salaries are usually below market. The gap is money the engineer gives up. Over a four-year vesting period, a sizable cut in pay each year adds up to a large sum the engineer gives up. Equity is not free in that bargain. The engineer is exchanging cash compensation for a speculative position in the company.

That commitment does not make the equity dependable income. At an early-stage AI startup, a large proportion of equity will ultimately be worth nothing, so mission and learning may provide stronger reasons to join than an expected payout. The uncertainty strengthens the case for scrutiny. It does not justify ignoring the terms because the outcome is unknowable.

A deep salary discount with a thin grant is especially hard to defend because the engineer gives up pay right away while any upside is still years away. An engineer can choose a risky opportunity with open eyes, but should know how the pay and equity work in time to make that choice. The useful test is an investment question. Would the engineer write a check equal to the salary forgone over four years to buy that equity position at the company’s current valuation?

Why pre-offer stage context changes what an engineer can actually do

That investment question belongs before the candidate is being asked to accept. When the engineer knows the stage context, they can judge whether the job is worth the pay they would give up, ask questions that fit where the company stands, and decide what pay and equity to expect. Timing changes the available choices. Information received with an offer can still help, but it arrives after weeks of interviews built on incomplete assumptions.

The same question also lands differently at different moments. Asking about funding stage before interviews is ordinary due diligence, while discovering that basic fact during negotiation can signal that the candidate has not prepared.

Speed makes that preparation more valuable. Senior ML searches can close quickly, and candidates may be weighing several offers and counter-offers at once. As Series A AI startups use cash to compete with large-tech counter-offers, equity can become a differentiating variable between packages. A candidate who understands it can compare offers without starting from scratch. Yet many job descriptions and recruiter messages omit stage, funding history, and ownership context, leaving candidates deep into a process before those comparisons become possible.

What stage-aware matching actually delivers that a job board cannot

A stage-aware talent network can use the company’s stage, recent funding, and typical equity range for the role as inputs to matching before the first screen. That changes the initial decision. The engineer can judge whether the job fits before there is a finished offer to weigh.

For ML candidates, that context also supports role validation. The company’s budget and headcount can help establish whether it is seeking a research engineer or applied ML work, and whether its usual equity range fits that category. The match becomes more specific than a shared job title. A candidate can spot a mismatch before spending time in technical interviews showing skills for a role with a different pay range.

The quality of the equity terms provides another filter. At seed stage, documented refresh and acceleration language gives candidates a firmer basis for comparison than generic option-grant language. The emerging talent-agent model goes further by representing engineers and negotiating packages proactively, rather than leaving every compensation question until the employer produces an offer.

The benefit depends on what the network actually knows. A stage label without role benchmarks or funding context does not solve the problem, and early matching does not replace scrutiny of the final terms. The useful capability is preparation at the point of introduction. For candidates weighing several opportunities at once, that preparation helps them decide faster without rushing to judge equity they know little about.

The due-diligence questions that stage context makes possible — and when to ask them

For engineers using this context, due diligence has two phases. Before the first conversation, the task is to establish whether the opportunity belongs in the right comparison set. The candidate should ask what funding stage the company is at and whether the stated equity range fits both that stage and the actual role category. Those questions establish the benchmark before either side anchors on a proposed grant. They also expose a role-title mismatch while it is still straightforward to discuss.

Recent funding should prompt more specific questions. Has the company raised recently enough that a 409A revaluation is likely before the grant vests, and does the strike price reflect current fair market value? Is the option pool sized before or after the most recent round, and is an expansion planned that would dilute the grant? These are questions about the starting position. They help establish what the advertised percentage describes, rather than treating it as a complete statement of ownership.

The engineer should ask about the refresh policy and its interval, because the absence of a documented policy is a warning sign. The post-termination exercise window deserves an explicit discussion, including whether an extension is negotiable. The engineer should ask whether vesting schedules are calendar-based or milestone-based and whether acceleration on acquisition is single-trigger or double-trigger. The candidate should also ask what the cap table looks like after the most recent round and what liquidation preferences sit ahead of common stock.

Alternative Minimum Tax exposure on ISOs needs to be modeled before the engineer commits cash. The broader purpose is not to become a cap-table expert, but to make the offer stage a confirmation of understood terms. If the engineer would not write a check equal to four years of forgone salary for that position at the current valuation, the response is to negotiate more salary, more equity, or both.

Sources

  1. CRV
  2. How AI is changing the compensation game for VC-backed startups
Filed underEquity Norms

More in Equity Norms