Simplify Logo
Perplexity AI

Perplexity AI

Advanced answer engine delivering source-backed replies

Member of Technical Staff - Software Engineer, Connector Platform

Full-Time
$220k - $405k/yr
Mid, Senior
Palo Alto, CA, USA+3 more

More locations: Seattle, WA, USA | San Francisco, CA, USA | New York, NY, USA

Hybrid

Hybrid work is required; the posting lists San Francisco, Palo Alto, Seattle, and New York City.

About the job

Requirements
  • Experience designing and building backend systems that run in production, typically 4+ years for mid-level and more for senior and staff roles.
  • Strong system design skills, with a track record of building efficient, reliable, and scalable architectures, ideally including API integration, gateway, or platform-style systems with many heterogeneous downstreams.
  • Strong proficiency in at least one backend language such as Python, Go, or Rust, and the ability to work effectively in a multi-language environment.
  • Hands-on experience with modern infrastructure, for example AWS, Kubernetes, and related cloud technologies.
  • Depth in at least one of OAuth and authorization protocols, API, connector or MCP-server development, schema and semantic modeling, or tooling and evaluation for LLM-based agents.
  • Comfort working in security-sensitive areas such as authentication, authorization, and credential isolation, and making pragmatic trade-offs between safety, simplicity, and velocity.
Responsibilities
  • Own the design and implementation of the connector runtime, the system that registers, hosts, and executes built-in connectors, hosted MCP servers, and CLI-backed tools behind a single agent-facing interface.
  • Build and extend the semantic layer, including tool and entity schemas, capability metadata, relationship modeling, and mechanisms for capturing and applying organization- and account-specific corrections and knowledge.
  • Design the tool-discovery and tool-selection surfaces that agents use to find the right connector and call it correctly, optimizing for model accuracy and context efficiency.
  • Make agent loops robust through structured results, partial-failure and retry semantics, idempotency, pagination, rate-limit handling, and observability into every tool call an agent makes.
  • Define authentication, authorization, and credential-isolation patterns for connectors, including OAuth flows, bring-your-own-key, and per-organization credential boundaries, partnering with Security and Backend Platform on defense-in-depth.
  • Build the connector onboarding path, including schemas, fixtures, and evaluation suites, so new connectors ship with measurable quality, and drive the evaluation metrics that determine whether a connector works inside agent loops.
  • Set the technical bar for connector reliability and operability through service-level agreements, observability, error-rate monitoring, and incident response for an always-on, high-fan-out integration surface.
  • Partner with product and artificial intelligence teams to define clear connector interfaces and integration patterns so new agent capabilities can reliably build on the shared platform.

About the company

Perplexity AI provides an answer engine that delivers precise, reliable responses to user questions by using up-to-date sources. It serves individuals who want quick answers and businesses that need detailed information, drawing on current data and source links to back its results. The product works by retrieving information from reputable sources, compiling concise answers, and presenting citations to ensure trustworthiness. It differentiates itself from competitors by prioritizing current, source-backed information and offering access that spans personal and enterprise use, potentially supported by subscriptions, advertising, and partnerships. The company's goal is to be a dependable tool for immediate, accurate information across a wide range of queries.

Company Size

1,001-5,000

Company Stage

Private

Total Funding

$1.8B

Headquarters

San Francisco, California

Founded

2022

Get referred to Perplexity AI

See people who can refer or advise you

Simplify Jobs

Simplify's Take

What believers are saying

  • Perplexity said it answers over 1.5 billion questions monthly, showing massive product pull.
  • September 2026 Crusoe deal expands training and inference capacity across GB300 NVL72 clusters.
  • Hybrid Compute and PII-Tracer strengthen privacy positioning for legal, enterprise, and regulated workflows.

What critics are saying

  • CNN sued Perplexity on May 28, 2026; NYT, Chicago Tribune, Reddit, and Dow Jones remain active.
  • Answer-engine economics depend on expensive GPU capacity and partner control, squeezing margins fast.
  • If courts restrict copyrighted-source ingestion, Perplexity's core search product loses retraining fuel.

What makes Perplexity AI unique

  • Perplexity Computer fuses search, agents, files, devices, and frontier models into one workflow.
  • Hybrid Compute on Apple silicon keeps sensitive steps local without restarting tasks.
  • Crusoe partnership on September 15, 2026 proves enterprise demand for managed low-latency inference.

Help us improve and share your feedback! Did you find this helpful?

Benefits

Health Insurance

Dental Insurance

Vision Insurance

401(k) Retirement Plan

Company Equity

Company News

Yahoo Finance
Sep 15th, 2026
Crusoe signs multiyear Perplexity cloud deal for Nvidia AI chips

Crusoe has signed a multiyear cloud agreement to supply Perplexity with Nvidia AI chips for model training and inference, Bloomberg reported Tuesday. The deal's financial terms and capacity were not disclosed. Under the agreement, Crusoe will manage the technology throughout its operating life. More than 1,800 Crusoe employees will receive access to Perplexity's Enterprise Max product, which provides AI-powered search tools. The contract adds to Crusoe's growing cloud business, which recently secured a five-year agreement with Jane Street worth roughly $13 billion. In June, Crusoe reported 4.9 GW of contracted AI infrastructure capacity across its data centre projects and cloud platform, with a development pipeline exceeding 40 GW. Crusoe opened a 400,000-square-foot manufacturing plant in Tulsa, Oklahoma, on Monday.

Crypto Briefing
Sep 15th, 2026
Crusoe signs multiyear cloud deal with Perplexity for AI models

Crusoe signs multiyear cloud deal with Perplexity for AI models. The AI infrastructure provider will host and manage Nvidia chips powering Perplexity's search models, adding to a growing roster of major cloud contracts an hour ago Sponsored: Vera - AI-powered prediction market intelligence, built for serious analysts Explore Vera Crusoe, the data center and AI cloud infrastructure company, has locked in a multiyear cloud capacity agreement with Perplexity AI to support the search startup's model training and inference workloads. The deal puts Crusoe in charge of hosting and managing Nvidia AI chips on Perplexity's behalf, a managed GPU arrangement that gives the AI search company another option beyond the hyperscaler giants it already relies on. For Crusoe, the partnership is another headline-grabbing contract in what's been a breakout stretch. Earlier in September, the company announced an approximately $13 billion five-year AI cloud contract with Jane Street, the quantitative trading firm. What the deal actually covers. Under the agreement, Crusoe will host Nvidia chips dedicated to Perplexity's AI operations and manage those chips over their full lifetime. Crusoe SVP Erwan Menard confirmed the company's role in overseeing the hardware on an ongoing basis, which means Perplexity is essentially outsourcing not just the compute but the operational headaches that come with running large GPU clusters. The deal also includes a reciprocal element: Perplexity will make its Enterprise Max product available to more than 1,800 Crusoe employees. AI, tech, and the markets they move - in one daily briefing. Daily. Free. Join 34,000+ readers across crypto, finance, and policy. Financial specifics like chip volume, total contract value, and precise timelines weren't disclosed. Perplexity's cloud diversification strategy. Perplexity has been deliberate about not putting all its eggs in one cloud basket. The company already has substantial commitments with AWS and struck a $750 million deal with Microsoft Azure reported back in January 2026. Adding Crusoe to the mix gives Perplexity a third major infrastructure partner, one that specializes in AI workloads rather than general-purpose cloud computing. Crusoe's rapid ascent in AI infrastructure. The firm has built out what it describes as nearly 5 GW of AI infrastructure capacity, with a pipeline exceeding 40 GW. The company originally gained attention for using stranded natural gas at oil wells to power Bitcoin mining operations, converting what would otherwise be flared emissions into productive compute. Disclosure: This article was edited by Editorial Team. For more information on how Crypto Briefing create and review content, see its Editorial Policy.

Android Authority
Sep 15th, 2026
You can now run Perplexity's local AI agent on your Windows PC.

You can now run Perplexity's local AI agent on your Windows PC. Your gaming GPU can now moonlight as an AI agent. Sep 15, 2026 - 8:11 AM ET Karandeep Singh / Android Authority Add Android Authority on Google: * Perplexity's Portable Computer can now run on Windows PCs with RTX GPUs. * It runs AI tasks locally, keeping sensitive files secure and saving token costs. * Portable Computer requires an RTX GPU with at least 24GB of VRAM. Perplexity has been rolling out feature after feature this year. After the agentic "Personal Computer," the company launched a new Portable Computer late last month for NVIDIA DGX Spark systems. Now, Perplexity and NVIDIA are finally making Portable Computer accessible to even more people. NVIDIA announced today that Perplexity's Portable Computer now works on Windows machines with compatible NVIDIA RTX GPUs. This means users no longer need a $5,000 NVIDIA DGX Spark setup to run Portable Computer. It also finally opens up access beyond DGX Spark and Linux-based systems. Perplexity Portable Computer runs completely on-device, with the AI models, tools, and the agent itself installed directly on the user's computer. This is useful for two main reasons: First, users can hand off tasks to the agentic AI without worrying about privacy because everything runs locally. Second, on-device tasks do not incur token costs, making them more cost-effective. For more advanced workflows, Portable Computer can still use cloud-based models. However, it asks for permission before offloading any tasks to more powerful cloud alternatives. Though Portable Computer supports Qwen 3.8 27B and a post-trained version of Qwen 3.8 called PPLX 27B, these models remain exclusive to NVIDIA DGX Spark systems. Users on Linux or Windows machines with RTX GPUs can only use Qwen 3.8 locally. However, Perplexity is working to add support for the NVIDIA Nemotron 3.5 Lightning model on both Linux and Windows. Portable Computer also requires an RTX GPU with at least 24GB of VRAM. That means you'll need an RTX 3090, 3090 Ti, 4090, or 5090 to run it, or a professional workstation GPU like the RTX Pro 4000, RTX A5000, or RTX Pro 6000. None of these GPUs are affordable, but they are considerably cheaper than a DGX Spark setup.

The Mirror Democrat and Savanna Times-Journal
Sep 15th, 2026
HP Redefines the Mobile Workstation for the Era of Agentic AI.

HP Redefines the Mobile Workstation for the Era of Agentic AI. * 4 hrs ago News Highlights: * The new HP ZBook Ultra G3a, a thin-and-light 16-inch mobile workstation, supports advanced workloads, including up to 300B-parameter LLMs locally, with up to 192GB unified memory and 160GB dedicated to the VRAM, enabling users to simultaneously run multi-agent and professional AI workflows. * Perplexity Computer comes preloaded, including supported integration for Autodesk Revit, helping professionals apply AI to defined tasks without starting from an empty prompt box. PALO ALTO, Calif., Sept. 15, 2026 (GLOBE NEWSWIRE) - HP Inc. today announced the HP ZBook Ultra G3a, a new mobile workstation designed from the ground up to serve as a personal AI factory for professionals. Powered by AMD and enhanced by Perplexity, the HP ZBook Ultra G3a 16" combines workstation-class performance, local AI capabilities, and intelligent workflow experiences for demanding professional environments. As organizations move from AI experimentation to production, professionals still face fragmented tools, disconnected applications, and complex setup requirements. HP is addressing that gap by bringing together AMD-powered local performance, Perplexity Computer to complete end-to-end work, and Autodesk(R) ii] application support in one mobile workstation built for professional use. "Professionals need AI experiences that are powerful, reliable, and integrated into the way they work," said Jim Nottingham, Senior Vice President and Division President, Advanced Compute and Solutions, HP Inc. "With the HP ZBook Ultra G3a 16", HP is bringing a local AI factory and workflow-ready tools together to help customers apply agentic AI to their most demanding work." Local AI performance, engineered with AMD HP and AMD co-engineered the HP ZBook Ultra G3a 16" to support large local AI models, AI agents, and professional applications in a thin-and-light mobile form factor. Powered by up to AMD Ryzen AI Max+ PRO 495 and up to 192 GB of unified memory, this single-chip system helps professionals design, build, analyze, and run AI locally without moving work to a larger system or relying on the cloud for every task. "Together, HP and AMD are ushering in the era of Personal AI, where extraordinary AI power moves from the data center into the hands of professionals. For years, the most powerful AI has lived in the cloud," said Jack Huynh, Senior Vice President and General Manager of the Computing and Graphics Group at AMD. "Now, with up to 192GB of unified memory and the ability to run models up to 300 billion parameters locally, that power is right beside you, working alongside you and your most demanding applications. The new ZBook shows what's possible when AI becomes truly personal, giving people the freedom to explore bigger ideas, solve harder problems, and bring their most ambitious ideas to life." This on-device approach can help organizations reduce cloud dependency and manage inference costs, while giving professionals more control over performance, latency, privacy, and where work runs. Agentic AI, ready for professional workflows The HP ZBook Ultra is engineered to run large local models, AI agents, and demanding professional applications at the same time. Ready-to-use workflow tools help customers apply AI to defined tasks without requiring teams to assemble hardware, models, connectors, prompts, and instructions on their own. "Professionals shouldn't have to build their own AI stack to put their workstation to work," said Ryan Foutty, Vice President of Business at Perplexity. "With HP, we're bringing them Portable Computer, a mode of Personal Computer that runs AI models, agents, and tools on the device. Work starts locally, and when a step needs frontier intelligence, users can choose to bring in a cloud model without starting over. Businesses get the benefits of local AI without giving up access to the cloud or control over when they use it." HP and Perplexity are enabling professionals to use AI more effectively with the applications they already rely on. Powered by Perplexity Portable Computer for Windows's MCP Connector to Autodesk Revit(R), customers can use local, on-device AI alongside their project data, keeping proprietary data on-device and minimizing cloud token costs. This helps professionals spend less time selecting tools, finding instructions, and writing prompts, and more time designing, creating, and solving problems. Additionally, Perplexity and HP have co-developed task-based Skills exclusively available to HP customers, designed to help users offload repetitive tasks and spend more time doing the work they enjoy. Mobile workstation performance for advanced workloads Built for all-day productivity on the go, the HP ZBook Ultra G3a 16" gives architects, engineers, designers, and technical teams workstation-class performance in a highly mobile design starting at 4.2 lbs. and 17.9 mm thin. A new thermal design delivers up to 81.8% higher TDP[iii], enabling up to 100W of performance for advanced workloads away from the desk. * Run complex AI models, agents, and advanced workloads simultaneously, including up to 300B-parameter LLMs locally, powered by AMD Radeon(TM) integrated graphics with up to 192GB unified memory, 160GB dedicated VRAM, and 55 TOPS NPU. * Support demanding design and visualization workflows, including Autodesk Revit 2027, Autodesk 3ds Max, and Chaos Enscape visualization. * Stay productive beyond Wi-Fi with secure 5G through HP Go connectivity. * Extend device life and simplify serviceability with an easily IT-replaceable keyboard and design features built for long-term use. A more controlled path for IT IT leaders are being asked to support AI adoption while managing cost, security, and governance. The HP ZBook Ultra G3a 16" is a new anchor in HP's broader agentic AI portfolio, alongside HP ZGX Nano, HP ZGX Fury, HP Boost, and HP Remote System Controller, giving IT teams a more controlled path to scale local AI. In addition, HP's new Local AI Value Calculator, available on HP.com will bring HP Z workstations customers a clear understanding of how moving AI workloads from the cloud to local devices can lower costs, provide valuable insight into performance, and help deliver faster return on investment for everyday AI use. By supporting local inference, large memory capacity, elastic scaling, and longer device life, HP gives organizations greater flexibility as AI models, frameworks, and workflows evolve. Pricing and availability The HP ZBook Ultra G3a 16" is expected to be available in October 2026. Pricing will be provided closer to availability. HP Inc. (NYSE:HPQ) is a global technology leader redefining the Future of Work. Operating in more than 180 countries, HP delivers innovative and AI-powered devices, software, services and subscriptions that drive business growth and professional fulfillment. For more information, please visit: HP.com. [i] Based on HP's internal analysis of competitive laptop PCs with AMD, Intel, Apple, and Qualcomm processors and configurable discrete or internal graphics as of August 2026. Most capable local agentic laptop PC based on the highest available graphics memory (VRAM) capacity among compared systems. Higher VRAM capacity may support larger local AI models and execution of multiple local AI agents, depending on configuration, model size, software, and workload. [ii] Autodesk, Revit, and 3ds Max are registered trademarks or trademarks of Autodesk, Inc., and/or its subsidiaries and/or affiliates in the USA and/or other countries. All other brand names, product names, or trademarks belong to their respective holders. [iii] Up to 81% (actual 81.8%) higher TDP compared to the ZBook Ultra G1a with a max TDP of 55W and ZBook Ultra G3a with a max TDP of 100W

Scribble
Sep 8th, 2026
Why are AI companies opening cafes? Inside the AI coffee shop boom.

Why are AI companies opening cafes? Inside the AI coffee shop boom. Tl;dr. Anthropic ran a Claude Café pop-up in New York's West Village that drew 5,000 visitors and 10 million social impressions in a single weekend. Perplexity opened a permanent coffee shop in Cheongdam-dong, one of the most expensive luxury districts in Seoul, home to Chanel, Hermès and Louis Vuitton flagships. Corgi, an AI insurance startup, runs a 24/7 café that's become an unofficial clubhouse for San Francisco's startup scene, and just signed a lease three blocks from Madison Square Park in New York, in a market renting at roughly $394 per square foot. Cursor did its own version in SF. AngelList's CEO tweeted a poll about turning his office lobby into a café, and two weeks later it existed. None of these companies sell coffee. All of them are spending real money to make you sit in a real chair, in real cities, holding a real cup. Here's why that's not a marketing afterthought, it's the whole point. The premise. Here's what's actually happening. A cluster of AI companies, most of them selling something you'll never physically touch, a chat interface, a code editor, an insurance policy processed by a model, are opening coffee shops. Not branded content about coffee. Actual, staffed, operating cafes, with human baristas, real espresso machines, and rent. Anthropic's Claude Café took over Air Mail's newsstand in New York's West Village for a week, part of its "Keep Thinking" campaign. Visitors got free coffee for downloading the Claude app, "thinking" caps, tote bags, and copies of CEO Dario Amodei's book. Lines snaked down the block. Perplexity opened Café Curious as a permanent location in Seoul's Cheongdam-dong, a neighbourhood so expensive it's nicknamed "the 1% of Korea." At checkout, a touchscreen asks whether you're a Perplexity Pro subscriber, a yes gets you half off, a no gets you a QR code for a free trial. Corgi, an AI-native insurance carrier, runs a 24-hour café in San Francisco that's become "founder heaven," and has since expanded to Atlanta and signed a New York lease steps from Madison Square Park. Why the most expensive real estate, specifically. Here's the part worth sitting with. These aren't cheap pop-ups in overlooked corners. Cheongdam-dong is one of the most expensive commercial districts in South Korea, the kind of address luxury fashion houses fight over. The Herald Square corridor Corgi just leased into averages roughly $394 per square foot. West Village storefronts don't come cheap either. That's not incidental. When your entire product lives inside a browser tab, the location of your one physical footprint becomes the loudest, most legible thing you can say about your ambitions. Renting space next to Chanel and Hermès says, wordlessly, that you belong in that conversation. A company that could run entirely from a server farm choosing the single most expensive square footage available is making a deliberate, expensive statement: Scribble is not a niche tool, Scribble is a serious, prestige brand, and Scribble is willing to pay luxury-tier rent to prove it before you've typed a single prompt. Why coffee, specifically. Coffee is doing very specific work here, and it's not random either. A coffee shop is the most universally legible "third place" there is, not home, not work, somewhere you go to sit, think, and be around other people without an agenda. AI, as a category, has a trust problem baked into its own success: it's fast, disembodied, and increasingly associated with "slop," the flood of low-effort, machine-generated content nobody asked for. A café is the exact opposite of that reputation. It's slow. It's staffed by actual humans. It smells like something. You have to physically go somewhere to experience it. Anthropic made this explicit. Its pop-up was framed as a "zero slop zone," pointedly timed right after a competitor's video model flooded feeds with exactly the kind of content that word describes. Free coffee, pen and paper, a real book, a human barista, this is a company using the most analog experience available to argue, without saying it outright, that its product is the thoughtful alternative to the flood. The real currency: trust you can feel. As brand strategist Karine Hsu put it, "with so many things happening in AI right now, people need to differentiate, because everything feels like a commodity." When every company claims to be the smartest, fastest, most capable model, those claims cancel each other out. A physical space doesn't. You can't fake standing in a room. You can't fake a barista handing you a cup. In a category where the product itself is often invisible and every competitor's marketing sounds identical, a real chair in a real city becomes rare, expensive, and durable proof that a company is willing to show up in a way its competitors haven't. It's marketing you can't screenshot and dismiss as another polished ad, because you were physically inside it. Why it works. Because it inverts what people expect from a company whose product is entirely digital. The instinct might be to assume a software company would express ambition through more software, a slicker interface, a flashier demo. Instead, the smartest of them are spending money on the most stubbornly physical, human thing available: a place to sit down. That inversion is what makes it newsworthy in the first place, Anthropic's café became a story precisely because nobody expected an AI lab to care about pen and paper and tote bags. It also converts a one-time visitor into a walking advertisement. Thousands of people posted photos of "thinking" caps arranged on coffee tables. That's a receipt, physical, ownable, sharable, in a way a downloaded app icon never is. Tips to steal. * Spend your physical footprint on prestige, not just proximity. Where you show up in real life says as much about your ambition as what you say. Choose the location like it's a headline. * Counter your category's biggest weakness with your physical space. If your industry has a trust or authenticity problem, build a space that argues the opposite through pure sensory experience, not slogans. * Give people a receipt they can hold. A cap, a tote, a cup, something physical people photograph and keep, travels further than a digital impression ever will. * Make the analog choice on purpose. Anthropic didn't hand out screens. It handed out pen and paper. In a category defined by digital output, going deliberately analog is the loudest thing you can do. * Know the difference between a stunt and infrastructure. A one-week pop-up builds buzz. Corgi's multi-city, always-open café builds a community that keeps coming back. Decide which one you're actually building. None of these companies needed a coffee shop to keep their servers running. They needed one to prove, in a way no benchmark score or demo video ever could, that a real person could sit in a real chair and feel something other than being sold to. When your whole product is invisible, the most expensive thing you can buy is proof that you're real. Frequently asked questions. Why are AI companies opening coffee shops?+ Which AI companies have opened cafes?+ Why do these companies choose expensive locations?+ What was Anthropic's Claude Café?+ Is this a real trend or a one-off?+