Software Engineer - Systems ML, Frameworks / Compilers / DL-Kernels
About the job
- Experience or ability to develop software for machine learning compiler frameworks, AI runtimes, deep learning kernels, or related tooling.
- Ability to analyze deep learning models and optimize them for specialized AI hardware.
- Ability to collaborate with AI researchers and hardware design teams.
- Knowledge of software development best practices, optimization, and performance tuning techniques.
- Develop software platform components for training and serving deep learning model architectures on specialized hardware.
- Work on PyTorch and generative artificial intelligence framework components, AI compilers and runtimes, high-performance deep learning kernels, or tooling for machine learning workloads.
- Analyze deep learning models and optimize them for MTIA hardware.
- Partner with hardware design teams to develop high-performance optimizations.
- Design features, optimize applications, and implement performance-tuning techniques using software development best practices.
- Develop machine learning compiler frameworks and contribute to next-generation hardware-software co-design for artificial intelligence domain-specific problems.
About the company
Meta builds Facebook, Instagram, WhatsApp, Messenger and Threads, apps that billions of people use to message friends, share photos and video, and follow creators and businesses. Most of its revenue comes from advertising: businesses buy placements across these apps, and Meta's AI systems decide who sees each ad and report the results. Its Meta Superintelligence Labs group builds the company's own AI models, the Meta AI assistant and the Muse agent app. Its Reality Labs unit makes Quest headsets and, with eyewear maker EssilorLuxottica, Ray-Ban and Oakley smart glasses. Meta was founded as Facebook in 2004 and is based in Menlo Park, California.
Company Size
10,001+
Company Stage
IPO
Headquarters
Menlo Park, California
Founded
2004
Get referred to Meta
See people who can refer or advise you
Simplify's Take
What believers are saying
- A booming ad business backs pay and stock: Meta's revenue rose 28% to $60.8 billion in the June 2026 quarter.
- AI engineers get vast resources: Meta expects $130 billion to $145 billion of capital spending in 2026.
- Wearables teams are scaling: in February 2026, Meta sought to double glasses output to 20 million units this year.
What critics are saying
- Job security is shaky: about 8,000 employees were hit by Meta's May 2026 layoffs, aimed at funding its AI push.
- Meta's biggest apps still face a breakup fight: the FTC appealed its Instagram and WhatsApp court win in January 2026.
- Hardware plans get reshuffled: Meta cut about 1,500 Reality Labs jobs in January 2026, shifting from the metaverse to wearables.
What makes Meta unique
- Meta's four apps form one huge audience: 3.56 billion people used them daily in March 2026.
- Meta leads smart glasses through eyewear giant EssilorLuxottica: their Ray-Ban and Oakley models sold over 7 million pairs in 2025.
- Meta builds consumer AI agents in-house: its Muse app, launched September 2026, topped Apple's US free-app chart within ten days.
Help us improve and share your feedback! Did you find this helpful?
Benefits
Stock Options
Company Equity
Mental Health Support
Flexible Work Hours
Growth & Insights and Company News
6 month growth
8%1 year growth
8%2 year growth
-1%Meta and Lighthouse Guild, a nonprofit serving the blind and low-vision community, sponsored a team of eight runners in the Army Ten-Miler in Washington, D.C., on October 11. The team included blind US Army veterans Roosevelt Morring, Rob Sanchas, and Shawntina Gibson. Each blind runner wore Oakley Meta Vanguard AI glasses and ran with a guide from Achilles International. Lighthouse Guild's president and chief executive officer, Thomas Panek, was guided by Meta's Kelly Ingham. The race highlighted Meta's pledge to give free Ray-Ban Meta AI glasses to every blind US veteran. More than 130,000 veterans are legally blind and eligible. "Technology is only useful if it reaches the people who need it," Ingham said.
Meta Platforms shares rose 2% on Tuesday morning after Truist raised its price target on the stock to $1,000 from $763. Truist analyst Youssef Squali cited the potential of Muse Wonder, Meta's newest artificial intelligence (AI) model, to strengthen the company's competitive position. The launch has drawn renewed attention from analysts. Meta is expanding its AI efforts as competition among major technology companies intensifies. Investors want to know whether new models can improve its products and support its wider business. Third-quarter results could show whether those investments are paying off.
Meta says it has reached the AI "frontier" with two new models, Muse Wonder and Muse Wonder Full, chief executive officer Mark Zuckerberg told the Financial Times. He also rejected calls from OpenAI, Anthropic, and others for the industry to "pace the frontier." Meta's own benchmarks show the models beating OpenAI's GPT-6 and Anthropic's Claude Fable 5.1 in areas such as coding and memory. Artificial Analysis called them "a material jump in intelligence over previous Meta models," and Vals ranked Wonder first on cybersecurity tasks. The models will power upgrades to the Muse assistant, which Meta plans to monetise through $20 and $100 monthly plans. The company plans $145 billion in capital spending on compute in 2026, up 101% from 2025.
Meta has launched "Your Algorithm" on Facebook, letting users shape what appears in Feed and Reels by describing their preferences in plain language. The company announced the AI-powered feature on Tuesday, after first rolling it out on Instagram. Changes show up in recommendations in real time. Facebook may ask follow-up questions to learn what users liked or disliked. Meta also updated the three-dot menu on posts. Users can open "Your Algorithm" there or choose "Interested" or "Not interested." Instagram introduced the feature in June, extending it across Feed and Explore after a Reels launch in December 2025. Threads has a related tool called "Dear Algo.
Meta's chief financial officer, Susan Li, warned executives that its new AI model could hurt the company financially if it failed, according to Alexandr Wang. Wang said on Stripe co-founder John Collison's Cheeky Pints podcast that Li pulled him aside "many times" at an executive offsite in Hawaii. "If it's good, it's all good," Wang recalled her saying. "But if it's bad, it's really, really, really bad." Meta this week launched its frontier models Muse Wonder and Muse Wonder Full from Meta Superintelligence Labs. Pivotal Research analyst Jeff Wlodarczak said they are at or near parity with leading labs, and Wonder Full scored 56 on the Artificial Analysis Index against 58 for Claude Opus 5.5. Wonder costs $4/$20 per million tokens. Wonder Full is sold only through Muse Code subscriptions at $50-$200 a month.