Full-Time
Applied AI Engineer
AI Platform
Zapier
1,001-5,000 employees
Automates cross-app workflows via Zaps
No salary listed
San Francisco, CA, USA + 2 more
More locations: Europe | North America
Remote
Remote work is available within the listed regions and eligible countries; Zapier maintains a country eligibility list.
 AI & Machine Learning (1) |
|---|
Get referred to Zapier
See people who can refer or advise you
- At least 4 years of experience in software engineering, including experience building and operating production AI/ML systems.
- At least 1 year of experience in LLM Ops, ML Ops, or adjacent platform/infrastructure work.
- Experience working through the full lifecycle of building, testing, deploying, and scaling machine learning or large language model architectures.
- Solid engineering fundamentals and comfort with typed languages and modern backend practices.
- An understanding of how to balance reliability, latency, cost, quality, and maintainability in production systems.
- Contribute to shared AI Platform capabilities that support teams building with machine learning and generative AI across Zapier.
- Help develop and maintain core services such as the LLM proxy server, platform APIs, and reusable tooling that standardize how teams access and operate models in production.
- Build and improve parts of the LLM Ops and ML Ops stack, including observability, monitoring, evaluation workflows, and operational tooling.
- Design and implement systems that improve the performance, reliability, safety, and cost efficiency of AI-powered experiences.
- Collaborate with engineers across product, infrastructure, and data teams to ensure AI components are reusable, well-documented, and easy to adopt company-wide.
- Evaluate emerging tools, models, and patterns in the AI ecosystem and help determine which should be incorporated into Zapier’s shared platform.
- Experience with TypeScript and Python.
- Experience in LLM Ops, ML Ops, or adjacent platform/infrastructure work.
- Experience contributing to backend systems, developer tooling, internal platforms, or infrastructure that supports other engineers.
Zapier is a software service that helps people connect different web apps so they can automate repetitive tasks. It works by letting users build
Company Size
1,001-5,000
Company Stage
Seed
Total Funding
$1.4M
Headquarters
San Francisco, California
Founded
2011
Get referred to Zapier
See people who can refer or advise you
Simplify's Take
What believers are saying
- ZoomInfo opened API access for every customer on July 27, 2026, expanding Zapier usage.
- August 2026 product updates added Tableau, Canva, BambooHR, Salesloft, and Looker.
- Zapier's beta MCP client links Claude and ChatGPT to live business actions.
What critics are saying
- Zapier's March 2026 layoff tracker shows 70 roles cut, signaling ongoing margin pressure.
- n8n self-hosting and Make's cheaper operations undercut Zapier on high-volume automations.
- OpenAI and Anthropic can route around Zapier connectors, eroding its distribution moat by 2027.
What makes Zapier unique
- Zapier MCP exposed 9,000+ apps and 30,000+ actions on August 17, 2026.
- Zapier still owns the easiest no-code workflow layer for nontechnical business teams.
- OAuth handling, rate limiting, and SOC 2 controls reduce customer IT burden.
Help us improve and share your feedback! Did you find this helpful?
Benefits
Work from anywhere
Competitive salary & bonus program
PTO
Health, dental, & vision coverage
Retirement plan with company match
Stock options
2 annual company retreats
Parental leave
Home office setup stipend
Professional development allowance
Growth & Insights and Company News
6 month growth
↑ 0%1 year growth
↑ 0%2 year growth
↑ 0%Five areas where SoftSolz falls short compared to competitors. Explore key gaps in SoftSolz compared to leading tools. Discover five areas where SoftSolz lags behind competitors like Zapier, QuickBooks, and Sprout Social, and learn when it's still the right choice. While SoftSolz offers a robust suite of features, there are specific areas where it doesn't quite measure up to some of its competitors. Understanding these gaps can help you make an informed decision about whether SoftSolz is the right tool for your business needs. Does SoftSolz have fewer automation connectors than Zapier? Yes, SoftSolz currently offers fewer automation connectors compared to Zapier. Zapier is renowned for its extensive library of over 3,000 app integrations, making it a powerhouse for businesses that rely heavily on automation. If your operations depend on connecting a wide array of apps seamlessly, Zapier might be the better choice. However, if your automation needs are more straightforward and you prefer a tool that integrates well with the core apps you use, SoftSolz can still be a viable option. Why does SoftSolz lack a general ledger? SoftSolz does not include a general ledger, which means businesses must continue using dedicated accounting software like QuickBooks or Xero for comprehensive financial management. These platforms offer robust accounting features that SoftSolz does not aim to replicate. For businesses that require detailed financial tracking and reporting, sticking with QuickBooks or Xero is advisable. SoftSolz remains a strong choice for those who need a versatile tool for project management and collaboration without the need for integrated accounting. How does SoftSolz compare to Sprout Social in social networks and analytics? SoftSolz offers fewer social network integrations and less analytics depth than Sprout Social. Sprout Social excels in providing detailed social media analytics and supports a broader range of social networks, making it ideal for businesses focused on social media marketing. Is SoftSolz lacking third-party security certification? Currently, SoftSolz does not have third-party security certifications, which can be a concern for businesses with stringent security requirements. Tools with certifications like ISO 27001 or SOC 2 provide an added layer of trust and compliance. For organizations where security is paramount, choosing a certified platform is recommended. SoftSolz is still a good fit for businesses that prioritize ease of use and flexibility over formal security certifications. Does SoftSolz have a smaller ecosystem of templates and integrations? Compared to long-established platforms, SoftSolz has a smaller ecosystem of templates and integrations. This can limit customization and scalability for businesses that rely on a wide range of pre-built solutions. For companies that need extensive customization options, platforms with a larger ecosystem might be more suitable. SoftSolz is ideal for those who prefer a simpler, more streamlined approach to project management. If any of these five areas are dealbreakers for your business, it might be wise to explore other options. However, if SoftSolz aligns with your core needs, its pricing and feature set can offer significant value. For more details, visit its pricing page. #softsolz #business tools #software comparison #automation #accounting
Relay AI shutdown sends founder back to Google Chrome. August 18, 2026 * The Relay AI shutdown ends customer access in September, while founder Jacob Bank returns to Google as Chrome's VP of Product. * The Relay AI shutdown gives Google workflow-automation talent as Chrome shifts from a browser into a potential agent workspace. * Relay tried to simplify repetitive business work with AI-powered automations, competing in a market long shaped by Zapier. * Google has not detailed Bank's product plans, but Chrome's Gemini integration offers an obvious foundation for more capable in-browser assistance. Table of contents. Relay AI shutdown marks the end of a familiar startup bet. The Relay AI shutdown is a tidy little story about an untidy market: a startup built to make work less repetitive is closing its doors, while its founder heads back to Google to help decide what work inside the browser looks like next. Relay will end access for paying customers on September 14, following the removal of free-tier access on August 15. Founder and CEO Jacob Bank said in a company announcement that Relay is winding down. Some employees are joining Google's Chrome organization with him. Bank's new LinkedIn role is VP of Product for Google Chrome, where he will oversee Chrome's product and developer-relations teams. For Relay customers, the immediate outcome is unglamorous: migrate workflows, export whatever can be exported, and find a replacement before the service goes dark. That's the less-discussed cost of the AI startup boom. Automation sounds wonderfully abstract until an approval chain, a content-review process, or a project-status update depends on a young vendor that runs out of road. Relay launched in 2021 with an ambition that was easy to understand. It wanted to be a more intelligent answer to Zapier, the long-standing default for connecting apps and shuttling data between them. Rather than asking users to build sprawling chains of triggers and actions, Relay emphasized templates and AI-assisted workflows for tasks such as drafting documents, editing copy, and coordinating project work. Relay ran into a brutal market. Incumbents such as Zapier, Microsoft, Salesforce, Atlassian, and Google now all pitch automation and generative AI as core product features. Meanwhile, customers have become more cautious about granting a small AI company broad access to internal documents, email, and business systems. A clever interface is not always enough when the buyer is thinking about compliance, reliability, and whether the tool will still exist next year. Why Google wants Relay's automation instincts. The Relay AI shutdown matters beyond one startup because Bank is not simply taking an advisory role at Google. He is stepping into one of the company's most strategically awkward jobs. Chrome has a vast user base, but the web browser is under pressure to become something more than a tab manager with a URL bar. Bank framed the move as a continuation of his career-long focus. "I've spent my whole career doing one thing: building tools that help people get more done with AI, without sacrificing their personal creativity or insights," he wrote on X. He added that Chrome is "a perfect place to collaborate with agents." That phrase - collaborate with agents - is doing a lot of work. Today, Gemini in Chrome can help summarize pages, answer questions about what is open in a user's tabs, and assist with writing. Gemini in Chrome is an optional in-browser assistant. The obvious next move is a browser assistant that can complete multi-step tasks: gather information from several sites, prepare a draft, update a service, and ask for approval at the moments that matter. Relay was chasing that same territory. The Relay AI shutdown removes a standalone contender, but it also delivers people who have spent years thinking about how to turn vague workplace intent into repeatable actions. Google has the distribution Relay never could. Chrome is already where a huge amount of office work happens, whether users are in Gmail, Docs, a CRM dashboard, or an internal web app. Still, browser agents have a trust problem that slick demos tend to skip. If an agent reads a page incorrectly, sends the wrong message, or clicks through a financial approval, it has done more than generate a slightly embarrassing paragraph. It has acted. Google will need clear permissions, visible previews, durable audit trails, and a dependable way for people to stop an agent mid-task. Frankly, that's harder than adding a chatbot sidebar. Jacob Bank's second Google chapter. Bank knows Google's productivity stack unusually well. His earlier startup, the scheduling app Timeful, was acquired by Google in 2015. He then spent more than six years at the company, including product leadership work across Gmail, Google Calendar, and Google Chat, before leaving to found Relay. That background makes the Relay AI shutdown feel less like a conventional acqui-hire and more like a boomerang return. Google gets an executive who understands both the company's internal machinery and the frustrations of trying to build workflow software outside it. Bank gets a chance to apply Relay's thesis at a scale that no independent productivity startup can realistically match. Google hasn't publicly laid out Bank's roadmap, and it would be a mistake to pretend SquaredTech know exactly which Relay ideas will show up in Chrome. Product integrations after talent deals are often less direct than headlines suggest. A startup's code may disappear; its lessons, design habits, and people can matter far more. Chrome's AI future could be useful - or exhausting. The Relay AI shutdown lands as Google pushes Gemini across Search, Workspace, Android, and Chrome. The company recently said Gemini had reached one billion users, a vast figure that reflects Google's ability to place AI in products people already use. But usage is not the same as affection, and it definitely is not proof that people want an agent involved in every browser session. My read is that Chrome has a real opening if Google keeps the work practical. Let an assistant compare open product pages, extract details from a tedious form, assemble research notes, or prepare a first-pass email from documents the user explicitly selects. Those are useful moments. An omnipresent assistant popping up to narrate the web would be exhausting. The Relay AI shutdown is disappointing for customers who took a chance on the product, but it also reveals where the AI automation battle is moving. The next winners may not sell standalone workflow tools at all. They may own the place where work begins - the browser - and persuade users to let software take the next click. Whether people actually grant that permission is the question Chrome now has to answer.
Connect sona.to to Zapier, Make and n8n. 06 Aug, 2026 There used to be one way to get a tool like sona.to into Zapier: the vendor built a connector, submitted it for review, and you waited. That is still how most integrations work, and it is why the app directory of any automation platform is a list of whoever had time to build one. Something changed quietly this year. Zapier, Make and n8n each added a client for the Model Context Protocol, the same standard Claude and ChatGPT use to connect to outside services. That means any product running an MCP server is reachable from all three today, with no connector and no review queue. sona.to runs one. Setting it up is the same idea everywhere. You create a token in the API page of your dashboard, add an MCP client step in your platform, paste the server address, and the available actions appear in a dropdown. The address is the same for everyone: https://mcp.sona.to/mcp In n8n it is the MCP Client Tool node. Add it under an AI Agent, set Endpoint to the address above, choose HTTP Streamable as the transport, and create a Bearer credential holding your token. The tool list fills in straight away, and you can run one from the node without building the rest of the workflow first. In Make it is the MCP Client module, using the Call a Tool action. Create a connection with the same address and token, and the Tool field turns into a dropdown of everything your token can reach. Make also parses the response into a structure you can map into the next module, so you are not pulling values out of a text blob. In Zapier it is MCP Client by Zapier. There are two pieces: a trigger that watches the available tools, and a Run tool action that actually calls one. Use the action, pick your tool, and it works like any other Zap step. It is still marked beta, so expect some rough edges. What you can do depends on the token. A read-only token can list your channels, read posts, pull analytics and look through SEO projects, issues and pages. A token with write access adds the ones that change something: create a post, run a site audit, apply an SEO fix. Creating a post is the interesting one, because it means a workflow can take a row from a spreadsheet or an approved item in your CMS and turn it into a scheduled post, without anyone opening sona.to. That covers one direction. For the other, add a webhook. sona.to sends a signed request when a post publishes, a post fails or a site audit finishes, and every one of these platforms can start a workflow from an incoming request. Put the two together and the loop closes: your workflow schedules the post, sona.to publishes it, and the same workflow hears back and logs it, notifies a channel, or updates a record. None of this costs extra. API access, the MCP server and webhooks are on every plan, including the free one. Create a token, paste one URL, and sona.to is a step in whatever you already run.
Zapier vs Make 2026: best fit for operators. Direct Answer Zapier vs Make for field service automation - both platforms offer powerful no-code workflows, but their pricing models, integration depth, and scalability vary significantly. For operators in 2026, Zapier is better for fast, linear automations; Make suits complex routing, data transformation, and high-volume operations. Key takeaways. * Zapier is ideal for simple, trigger-to-action workflows, with a task-based pricing model that's easier to predict for low-volume users. * Make excels in complex automation involving branching, filtering, and data aggregation, especially when developers need control or custom integrations. * For field service tasks like dispatching and notifications, Zapier offers faster setup and easier maintenance for non-technical teams. * Make is preferred when total cost of ownership matters, particularly for high-frequency operations, due to its operation-based billing. * Operators choosing between platforms should consider who will own the automation day-to-day - non-tech users lean toward Zapier, while technical roles often prefer Make. Why this matters. As field service workflows grow more complex and data-driven, operators and founders need tools that scale without sacrificing speed or control. In 2026, Zapier and Make represent two distinct approaches to automation: one focused on ease-of-use and broad reach, the other on deep customization and operational efficiency. The right choice depends heavily on how workflows will be maintained, the frequency of operations, and the team's technical familiarity with automation tools. For instance, if an operator wants a quick handoff from dispatch to customer SMS updates, Zapier likely wins due to its simplicity and robust connector library. However, for a system that filters technician availability, routes based on skill sets, and logs complex work orders, Make's module-based approach is more suitable. Additionally, 2026 has seen increasing demand for AI-augmented workflows in field service environments. Both platforms have adapted to support these new capabilities, though in different ways. Zapier continues to integrate native AI features that require less configuration, while Make provides modular tools that allow developers to incorporate AI into custom logic with greater flexibility. What changed. In 2026, both platforms have evolved their offerings significantly In 2026, many field teams are evaluating automation based on total cost at scale, which favors Make for high-volume scenarios. However, initial setup time and ease of maintenance still strongly favor Zapier for simpler use cases. Zapier has made strides in improving its connector library, especially for field service-specific platforms like ServiceNow and Salesforce. Meanwhile, Make's strength lies in how it handles nested logic and complex data transformations - capabilities that are more critical in environments where technicians must make quick decisions based on dynamic input. One key shift in 2026 is how teams evaluate long-term workflow fit and implementation tradeoffs. For example, a company with limited IT resources may find Zapier's intuitive interface and out-of-the-box integrations sufficient for basic needs. In contrast, organizations that rely heavily on scheduling logic, skill-based routing, or multi-step validation might find Make more aligned with their operational goals. * Zapier introduced pay-per-task billing, allowing teams to manage costs more precisely. This change allows users to better budget for workflows that may spike in volume without overcommitting to fixed plans. * Make continues to refine its credit system, offering better integration with APIs and support for AI features. This evolution reflects the growing need for granular tracking of automation usage across complex deployments. * The rise of AI-augmented workflows has added new dimensions to the comparison - Zapier's native AI tools are often more accessible, while Make offers modular AI use cases that can be integrated into more advanced automations. Recommended actions. * If your team is new to automation and needs fast, reliable workflows like dispatch notifications or status updates, start with Zapier. * For complex routing, data transformation, and high-frequency operations, make is the better long-term bet even if it costs more upfront. * Test both platforms side-by-side with a few real-world field service use cases to understand which fits your day-to-day maintenance needs and cost expectations. Frequently asked questions. Which is better for field service automation in 2026? Zapier is better suited for fast, linear workflows like SMS alerts and status updates. Make excels where workflows branch, loop, or require data manipulation before action, such as dynamic routing or technician scheduling. How do Zapier and Make handle pricing? Zapier charges per task in tiered plans, with pay-per-task available. Make uses a credit model that tracks each operation run, allowing more granular cost tracking for heavy users. Can I switch between Zapier and Make later? Yes, switching is possible, but it may require reconfiguring workflows and can involve downtime or data migration delays. Is Make harder to learn than Zapier? Yes. Make's flexibility comes with a steeper learning curve, especially for non-technical users who need full control. Zapier remains more approachable for quick deployments. Sources and evidence. * Make vs. Zapier: I Dragged and Zapped Until One Won Me Over explains both platforms' core differences in pricing and use cases * Zapier vs Make: Which Is Better? (2026) compares ease-of-use, integrations, and scalability in field service settings * Make vs Zapier: How Are Elevation Engine Different? Make outlines Make's credit system and custom integration support
Domains.co.za Launches Self-Hosted n8n VPS Hosting for AI and Workflow Automation. August 3, 2026 The offering is designed to provide organizations with greater control over their automation environments while supporting unlimited workflows and enterprise-grade performance. Domains.co.za has launched a new self-hosted n8n Virtual Private Server (VPS) hosting solution in South Africa, enabling businesses and developers to deploy and manage workflow automation and AI-powered applications on locally hosted infrastructure. The offering is designed to provide organizations with greater control over their automation environments while supporting unlimited workflows and enterprise-grade performance. The hosting platform comes with pre-installed n8n, the open-source workflow automation tool that enables users to connect applications, systems, and data sources through APIs and visual workflows. By offering a self-hosted deployment model, Domains.co.za aims to address growing demand for enhanced data privacy, regulatory compliance, and customizable automation environments. Unlike cloud-based automation platforms that often impose execution or workflow limits, the self-hosted VPS solution allows customers to run unlimited workflows while retaining full ownership and control of their data. The platform is positioned as a suitable option for businesses seeking to automate operations, orchestrate AI agents, and integrate a wide range of business applications. The company highlighted that hosting n8n on a dedicated VPS provides a stable, always-on environment capable of supporting real-time workflow execution. Businesses can benefit from continuous automation, predictable performance through dedicated computing resources, enhanced security controls, and the flexibility to scale infrastructure as automation requirements grow. To ensure service reliability, Domains.co.za hosts the solution within the Teraco data centre in South Africa, which offers resilient power infrastructure designed to maintain operations during load-shedding. The company says the facility provides high availability through battery backup systems and diesel generators, supporting uninterrupted workflow execution for mission-critical applications. The platform supports integrations with hundreds of third-party services, including business productivity, CRM, collaboration, e-commerce, and AI platforms such as Google Workspace, Salesforce, HubSpot, Slack, Shopify, and OpenAI. Users can build workflows ranging from simple business process automations to complex AI-driven orchestration pipelines using either a visual interface or custom JavaScript and Python code. Domains.co.za also positions the offering as a locally hosted alternative to cloud-based automation platforms such as Zapier and Make. According to the company, the self-hosted model offers flat-rate VPS pricing, unlimited workflow execution, local data residency, and greater control over compliance requirements, including support for organizations seeking to align with South Africa's Protection of Personal Information Act (POPIA). With the launch, Domains.co.za is expanding its infrastructure portfolio to support the growing adoption of AI, workflow automation, and digital transformation initiatives among businesses seeking locally hosted, scalable, and privacy-focused automation solutions. The TechAfrica News Podcast