Full-Time
Member of Technical Staff
Android
Mercor
1,001-5,000 employees
Automates candidate screening and matching
Compensation Overview
$130k - $400k/yr
Company Historically Provides H1B Sponsorship
San Francisco, CA, USA
In Person
Five days per week in the San Francisco office required.
 Software Engineering (1) |
|---|
Get referred to Mercor
See people who can refer or advise you
- At least 5 years of professional Android software engineering experience.
- Strong expertise in Kotlin and Jetpack Compose.
- Experience shipping and maintaining production Android applications through the Google Play Store.
- Strong understanding of modern Android architecture, testing, debugging, and performance optimization.
- Demonstrated experience making architectural decisions and driving technical direction across projects.
- Fluent use of modern artificial intelligence software development tools such as Cursor, Claude Code, GitHub Copilot, or ChatGPT to improve engineering productivity.
- Strong software engineering fundamentals and product judgment.
- Experience mentoring engineers and raising the technical bar across a team.
- Strong communication skills and a bias toward shipping.
- Lead the architecture and development of Mercor's native Android application.
- Design, build, and ship high-quality Android applications using Kotlin and Jetpack Compose.
- Own major product initiatives end-to-end, from technical strategy and architecture through production launch.
- Define Android engineering standards, architecture patterns, testing strategy, and release quality.
- Collaborate with Product, Design, AI, Backend, and Infrastructure teams to deliver customer-facing experiences.
- Leverage modern artificial intelligence development tools to accelerate engineering while maintaining a high quality bar.
- Optimize application performance, reliability, observability, and long-term maintainability.
- Mentor engineers and help establish a strong mobile engineering culture.
Mercor automates the recruitment process by using artificial intelligence to match job seekers with employers. Candidates upload their resumes and complete a 20-minute AI-led video interview, which a large language model analyzes to create a detailed profile of their skills and predicted performance. Unlike traditional recruiters that rely on manual screening, Mercor uses these automated interviews and data-driven matching to vet hundreds of thousands of candidates across various industries simultaneously. The company's goal is to reduce human bias and speed up hiring by providing employers with a pre-vetted pool of talent through a fully automated platform.
Company Size
1,001-5,000
Company Stage
Series C
Total Funding
$483.6M
Headquarters
Menlo Park, California
Founded
2023
Get referred to Mercor
See people who can refer or advise you
Simplify's Take
What believers are saying
- Mercor said June 2026 annualized gross revenue hit $2 billion, doubling in four months.
- Fortune reported OpenAI, Anthropic, Meta, and Google among Mercor's named customers.
- Mercor's July 2026 Deeptune deal strengthens agent-training products alongside expert evaluation.
What critics are saying
- March 2026 LiteLLM breach triggered class actions over passport scans, biometrics, and SSNs.
- Mercor still faces reputational damage if customers distrust contractor data handling by Q4 2026.
- Revenue depends on frontier labs; if OpenAI or Anthropic internalize evals, Mercor collapses.
What makes Mercor unique
- Mercor's human-expert marketplace trains frontier models with domain specialists, not generic recruiters.
- Deeptune acquisition on July 9, 2026 adds simulation environments for agent training.
- Mercor's 5,001-employee network and 15,932 job postings create dense talent supply.
Help us improve and share your feedback! Did you find this helpful?
Benefits
A $20K relocation bonus (if moving to the Bay Area)
A $10K housing bonus (if you live within 0.5 miles of our office)
A $1K monthly stipend for meals
Free Equinox membership
Generous equity grant
Performance bonus
Health insurance
Company News
What are AI training jobs? A guide for PhDs and master's graduates. 07/18/2026 What is an AI training job? AI training jobs involve helping improve artificial intelligence systems by providing human feedback on model outputs. Modern AI systems are trained on enormous amounts of data, but they continue to rely on human evaluation throughout their development. Companies use AI trainers to assess the quality of responses, identify factual errors, evaluate reasoning, test new capabilities, and generate high-quality training data. This feedback helps developers refine models, improve performance, and better align AI systems with human expectations. Depending on the project, AI training work may include: * Comparing multiple AI-generated responses and ranking which one is better. * Evaluating responses for factual accuracy, reasoning, clarity, or completeness. * Writing prompts that test an AI model's capabilities. * Revising or rewriting responses to create higher-quality examples. * Reviewing outputs within a specialized field such as medicine, law, software engineering, finance, biology, or education. * Identifying unsafe, misleading, or biased responses. The specific responsibilities vary by company and project, but the common goal is the same: providing structured human feedback that helps improve the quality, reliability, and usefulness of AI systems. If you're interested in exploring opportunities to contribute to AI development, After Your PhD partners with Mercor and Handshake to share AI-related job postings and opportunities. These roles can vary widely depending on the company, project, and area of expertise, but many involve using your academic background to help train, evaluate, and improve AI systems. What does an AI trainer actually do? The exact responsibilities depend on the company and the project, but most AI training work falls into a few common categories. One of the most common tasks is evaluating AI-generated responses. You may be presented with two answers to the same question and asked which one is better, then explain your reasoning using a detailed rubric. Your feedback helps companies understand what makes a response useful, accurate, and easy to understand. Other projects focus on writing prompts that challenge AI systems with difficult or realistic scenarios. These prompts help expose weaknesses in the model's reasoning and create better training data for future improvements. Some assignments involve fact-checking AI-generated content, identifying hallucinations, correcting technical mistakes, or rewriting weak responses so the model has a stronger example to learn from. For people with specialized expertise, the work becomes even more interesting. A biologist might evaluate scientific explanations, a lawyer could review legal reasoning, an economist may analyze financial responses, and a software engineer might assess AI-generated code. In these projects, you're applying your professional knowledge to improve how AI performs within your discipline. The benefits of AI training jobs. Flexible work That fits around your schedule. One of the biggest reasons AI training has become popular among academics is its flexibility and earning potential. Most projects are completed remotely, and many allow you to choose when you work. That makes these opportunities especially attractive if you're finishing your dissertation, teaching as an adjunct, interviewing for industry positions, or simply looking for additional income between jobs. While every person's experience is different, I've found that flexibility is one of the benefits people mention most often. Unlike many traditional part-time jobs, AI training often allows you to work when it's convenient for you instead of committing to fixed shifts. You continue using your academic skills. Many academics worry that leaving higher education means leaving behind years of specialized training. AI training is one of the few opportunities where that's not necessarily true. Rather than stepping away from research, writing, analysis, and critical thinking, you're applying those same skills in a completely different setting. Evaluating evidence, explaining your reasoning, identifying errors, and communicating clearly are all valuable parts of many AI training projects. One benefit I didn't fully appreciate until talking with people working in AI training is how much they learned about AI itself. After spending weeks or months evaluating model outputs, many people said they became significantly better at using tools like ChatGPT in their own work. You begin to understand where AI excels, where it struggles, and how better prompts often lead to dramatically better results. Whether you're interested in research, consulting, teaching, or simply becoming a more effective AI user, that's a valuable skill to develop. Your expertise can lead to higher-paying opportunities. Not every AI training project pays the same. Generalist projects typically pay less than projects requiring specialized expertise. If you have experience in medicine, law, software engineering, finance, education, science, or another technical discipline, you may qualify for projects that pay significantly more than entry-level opportunities. While compensation varies considerably by company and project, subject-matter expertise is often one of the biggest factors influencing pay. The downsides. Like any type of freelance work, AI training isn't perfect. AI is also changing the job market. One reality that's difficult to ignore is that many of the same advances creating AI training jobs are also changing the broader job market. As AI systems become more capable, some tasks that were previously performed by humans are becoming automated, and organizations are rethinking how certain types of work are completed. Exactly how AI will affect employment over the long term is still uncertain, and the impact is likely to vary widely by industry and occupation. Rather than replacing entire professions overnight, AI is often changing specific tasks within existing roles. For many PhDs, this creates an interesting trade-off. By working in AI training, you're contributing to the development of technologies that may reshape parts of the labor market, while also gaining firsthand experience with the tools that are driving those changes. Whether you view that as an opportunity, a concern, or a combination of both is ultimately a personal decision. Income can be unpredictable. Perhaps the biggest downside is that most AI training work is project-based. Projects can begin and end with little notice, available hours may fluctuate, and there may be periods where little work is available. Because of this, I generally encourage people to think of AI training as supplemental income rather than something to rely on as their only source of financial stability. Not every project will be exciting. It's also important to have realistic expectations. AI training work isn't always intellectually stimulating. Some projects involve evaluating hundreds of similar responses or following detailed rubrics for long periods of time. That repetition can become mentally draining, especially if you're used to the variety and independence that comes with academic research. On the other hand, not every project feels this way. Many people enjoy assignments that closely align with their area of expertise, particularly when they're evaluating technical content or solving complex problems. Like most freelance work, the experience often depends on the specific project you're assigned. Not necessarily. While software engineering projects certainly exist, many AI training opportunities don't require programming experience at all. Companies are looking for experts across dozens of disciplines, including biology, chemistry, education, psychology, history, law, medicine, business, finance, writing, and languages. In many cases, your ability to evaluate information critically is far more valuable than your ability to write code. One unexpected benefit of working in AI training is that you begin to understand how large language models think, where they excel, and where they still struggle. If you're interested in learning more about artificial intelligence, these books are excellent places to start. A practical guide to using AI effectively in your everyday work. This is one of my favorite introductory books for professionals who want to understand tools like ChatGPT without diving into technical machine learning concepts. An accessible look at how artificial intelligence is changing work, business, and society. Rather than focusing on the technology itself, the book explores the broader implications of increasingly capable AI systems. If you're interested in understanding how modern AI systems are built and deployed, this book provides an excellent overview of large language models, evaluation, and production AI systems. A fascinating exploration of how researchers are trying to ensure AI systems behave the way humans intend. Topics like reinforcement learning from human feedback (RLHF), bias, and model evaluation connect directly to many AI training jobs. A balanced introduction to what AI can - and can't - do today. This is an excellent starting point if you're completely new to artificial intelligence. Ryan Collins PhD is an SEO Strategist at Go Fish Digital. Ryan completed his PhD in Media Arts and Sciences at Indiana University Bloomington in 2021. During his time at Indiana University, Ryan eventually pivoted into a career in SEO and Digital Marketing after having informational interviews with working professionals in SEO, working on side projects, and gaining industry experience.
White-collar professionals are earning up to $200 an hour training AI models, but workers say the reality falls short of expectations. Mercor, a San Francisco startup selling training data to AI labs, pays its 30,000 contractors upward of $4 million daily and was reportedly valued near $20 billion in early talks last July. Whilst some roles offer $350 an hour for psychiatry experts or up to $1,000 for venture capital partners, most positions average around $105 an hour. Workers report tight deadlines, flat-rate projects consuming more time than paid for, and inconsistent feedback. A deeper issue emerges as AI models quickly learn from expert corrections, causing work in specific specialties to dry up within weeks. Several contractors have sued data-training firms over alleged misclassification and underpayment. Economists debate whether this represents sustainable employment or a temporary phenomenon that will shrink as systems improve.
Catchup with JANUS: a skeptical audit of Salesforce's agents, Cognition goes vertical, and Mercor doubles to $20B. Analysts checked Salesforce's customers and found Agentforce struggling to win them over even as it crosses $1B in recurring revenue, and Microsoft turned agent hosting into a commodity cloud service. The gap this week is between how fast agent platforms are shipping and how convinced buyers actually are. Welcome to Catchup with JANUS, a periodic note from the iAgentic team on what's moving in agent orchestration. Platform announcements, security incidents, enterprise deployments, the developer-tools fault lines, and the things Iagentic think enterprise buyers should actually care about. No vendor spin. For weeks these notes have followed two long threads: a security story about how agents can be turned against you, and a platform story about incumbents racing to control how agents get built and governed. This week the security thread went quiet, and the platform thread hit its first real audit. The signal turned commercial: shipping is fast, and buyers are starting to ask whether any of it works. Headline moves. Salesforce. This is the next chapter of the Summer '26 story Iagentic has tracked since Multi-Agent Orchestration and the Agent Fabric control plane went generally available in June. On July 6, Salesforce made its Shopper, Buyer and Merchant commerce agents generally available and launched a Help Agent priced per resolution. Then on July 15, KeyBanc analysts said their customer checks came back weak, reporting that Agentforce "isn't winning over clients" even as its annual recurring revenue passes $1B. The "can they prove it works" question Iagentic raised when the control plane shipped now has a number attached to it. Cognition. After the $1B raise and the Devin Desktop pivot covered in earlier issues, Devin's maker went vertical. It shipped Devin Security Swarm on July 1, an agent that finds, validates and fixes vulnerabilities and topped a 50-item benchmark at lower cost per finding, then followed with SWE-1.7 on July 9, an in-house model posting near-frontier coding scores at roughly 1,000 tokens per second. The trajectory is consistent: own the interface, then the model, then specialized products on top. Mercor. Two weeks after the breach notifications and lawsuits Iagentic reported, the money has not slowed. Mercor is in talks to raise $500M at a $20B valuation, double its September mark, citing $2B in annualized June revenue, and is acquiring the agent-training firm Deeptune. A security incident that would sink a slower company has, so far, been a footnote to its growth. Microsoft. Advancing the Agent 365 governance push from earlier briefings, Foundry Agent Service hosted agents reached general availability: a framework-agnostic managed runtime that hosts agents built on LangGraph, the Copilot SDK or Microsoft's Agent Framework, with sandboxed sessions, durable state and scheduled routines. Microsoft is now in the business of running other people's agents, not just governing them. LangChain. Agent memory became a product. After a July 2 cluster of releases, LangChain launched OpenWiki Brains on July 10: general-purpose Markdown "wiki memory" that agents maintain themselves from Gmail, Notion, git and web sources. It continues the pattern of the leading open framework moving up the stack. Glean also shipped its July Drop of new content formats and governed workflows. Where the landscape is shifting. The clearest shift is that agent hosting is becoming a commodity cloud service. With Foundry Agent Service generally available, a major cloud will run your agents regardless of the framework that built them, which absorbs part of what standalone platforms used to sell. LangChain is pushing the other way, turning an open framework upward into memory and orchestration. The middle of the stack is getting squeezed from both ends. Pricing is moving too. Salesforce's pay-per-resolution model, arriving right after the same company spent June building a cross-vendor control plane, sets an expectation buyers will carry into every agent conversation: pay for outcomes, not seats. Easy to announce, hard to sustain, but once a large vendor offers it the question spreads. And the KeyBanc checks matter beyond Salesforce. Through June, the platform story was incumbents claiming the control-plane and governance narrative faster than they could prove it. KeyBanc is the first widely reported, demand-side read on whether those claims hold at scale, and at $1B in recurring revenue the answer was lukewarm. The velocity of announcements has outrun the evidence of adoption, and analysts are now measuring the difference. Where the openings are. Three places worth attention if you are evaluating orchestration vendors right now: * Ask for evidence, not demos. The KeyBanc read suggests headline revenue and reference logos can mask weak adoption. Ask for outcome data from customers who look like you, not a staged walkthrough. * Test the pricing claim. Pay-per-resolution sounds buyer-friendly. Check what counts as a "resolution," who adjudicates it, and whether the effective price at your volume beats a seat model. * Confirm framework portability. Managed runtimes now host multiple frameworks. If avoiding lock-in matters, verify your agents, memory and state can move between a hyperscaler runtime and your own without a rebuild. On the calendar. * Mercor's $500M round is expected to close this month: watch whether the pending lawsuits affect terms. * Salesforce commerce agents gain native ChatGPT and Gemini app integration: a template for distributing agents through consumer AI surfaces. * IBM holds a July 23 webinar on scaling agents across frameworks and clouds, which may finally date its remaining control-plane modules, still open from earlier briefings. * Still open: Databricks Omnigent adoption, and whether Sentry or MCP vendors ship a root-cause fix for the Agentjacking attack class. From the community. A quiet week for developer threads, but the governance-lag theme from prior issues is intensifying in analyst coverage: Gartner now projects 40% of enterprise applications will embed agents by year-end, up from under 5% in 2025. The Agentjacking storyline that dominated recent weeks is fading from front pages without a fix shipping, which reads as quiet risk rather than resolution. Early reaction to LangChain's OpenWiki is positive, with one recurring question: who maintains an agent's memory after go-live. Commentary points the same way, with Ethan Mollick arguing the real skill is now managing agents rather than prompting them. The through-line holds: developers want capability that gets out of the way; buyers want systems that refuse to act until an outcome is verified.
Mercor eyes $20B - read the AI 'revenue' number first. Mercor is raising at a $20B valuation on a $2B revenue run-rate. Here's how to read AI vendor numbers before you bet your stack on them. An AI company most of your customers have never heard of is about to be worth $20 billion. Mercor - a marketplace that supplies human experts to train and evaluate the frontier models everyone else builds on - is in talks to raise at a $20 billion valuation, roughly double where it sat nine months ago. The number worth studying isn't the valuation. It's the "$2 billion in revenue" underneath it. What actually happened. Per TechCrunch, Mercor is in early talks for a new round at a $20 billion valuation - up from $10 billion when it raised $350 million in October 2025 - and has reportedly already received a term sheet. CEO Brendan Foody says the company crossed $2 billion in annualized revenue in June, a 100% jump in four months. Alongside the raise, Mercor said it's acquiring Deeptune, an AI-agent-training startup whose entire team is joining. Forbes and Bloomberg reported the same valuation talks the same day. Now read the revenue line carefully. "$2 billion annualized" is a run-rate - a recent month multiplied out, not $2B booked over a year. And Mercor's model is a marketplace: it connects thousands of contract experts to AI labs that need training data and evaluations, which means a large share of that top line is pass-through - money that flows straight to the people doing the work, not margin Mercor keeps. (The exact take-rate isn't disclosed.) None of that makes Mercor a bad business. It makes the headline something other than what "$2 billion in revenue" sounds like. Why it matters for your business. You're going to be sold on numbers like these. Every AI vendor pitching you - the model provider, the agent platform, the tool your SaaS just bolted on - will wave a valuation and a run-rate as proof they're safe to build on. Those aren't the same claim. A doubling valuation on an annualized top line, in a business where most of the money passes through to contractors, tells you the category is hot. It tells you very little about whether that vendor will still exist, at that price, in three years. Its rule when Rushcommerce help a client pick an AI vendor: ignore the headline, read the unit economics. Ask what the revenue number actually measures - booked or run-rate, gross or net - where the margin lives, and what happens to your workflow if the vendor gets acquired, repriced, or runs out of runway. Mercor's rise makes a quieter point too: the "intelligence" in the models you rent rests on an army of paid human experts you never see. That's the real cost structure of AI, and it's worth remembering the next time a demo makes it look like magic. Key takeaways * Mercor is in talks to raise at a $20B valuation - double its $10B mark from October 2025 - and says it hit a $2B annualized revenue run-rate in June * "$2B revenue" is a run-rate on a marketplace top line; much of it is pass-through to the contract experts who do the labeling and evaluation, not margin * Headline valuation and run-rate are hype signals, not durability signals - judge AI vendors on unit economics and on what happens to you if they're acquired or repriced * Frontier models run on human expert labor you don't see - that's the actual cost base of the AI you rent Picking AI vendors to build on? Rushcommerce help small teams choose tools by durability and real economics, not funding-round theater - and Rushcommerce build systems you own, so a repricing is a memo, not an emergency. See how Rushcommerce work or get a vendor-durability review. * #mercor * #ai-funding * #vendor-risk * #ai-economics * #data-labeling Tommy Rush - Founder, Rush Commerce Operator turned builder. 15+ years running operations - now shipping the systems businesses run on. More Get The Rush Report weekly - one email, zero fluff.
AI training startup Mercor has raised $350 million in Series C funding led by Felicis Ventures, with participation from Benchmark, General Catalyst, and Robinhood Ventures. The round values the company at $10 billion, a fivefold increase from its February 2025 valuation. Founded in 2023, Mercor connects AI labs with domain experts in fields like law, finance, and medicine to train AI models. The company's valuation has grown from $2 billion in late 2024 to $10 billion in just over a year. Mercor's clients include OpenAI and Anthropic. The startup now employs over 300 people globally. Its rapid growth reflects increasing demand for human expertise in AI model training.