{"id":1143797,"date":"2025-07-03T03:39:40","date_gmt":"2025-07-03T10:39:40","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-job-opportunity&p=1143797"},"modified":"2026-04-01T08:00:22","modified_gmt":"2026-04-01T15:00:22","slug":"research-intern-for-machine-learning-group-embodied-ai-research","status":"publish","type":"msr-job-opportunity","link":"https:\/\/www.microsoft.com\/en-us\/research\/opportunity\/research-intern-for-machine-learning-group-embodied-ai-research\/","title":{"rendered":"Research Intern for Machine Learning Group – Embodied AI research"},"content":{"rendered":"\n\n\n

Position Title:<\/strong> Research Intern for Machine Learning Group – Embodied AI research
Job Type:<\/strong> Full-time Intern
Number of Openings:<\/strong> 2-4
Location:<\/strong> Beijing<\/p>\n\n\n\n

Group Introduction:<\/strong><\/p>\n\n\n\n

We focus on building foundation models for embodied AI. In terms of algorithmic research, our interests include but are not limited to latent action learning, VLA, World Model, RL, and other research directions. We focus on learning decision-making knowledge from internet-scale video data to enhance the generalization capabilities of embodied AI foundation models. Over the past few years, we have published numerous papers in top international conferences and journals. Our proposed IGOR framework achieves, for the first time, the transfer of human movements to robots solely through latent actions learned from internet-scale videos, opening new possibilities for human-to-robot knowledge transfer and control. We also have extensive experience in RL; our Mahjong AI system Suphx became the first AI system to achieve the rank of ten-dan on the internationally renowned professional Mahjong platform “Tenhou,” surpassing the average level of top human players in the platform’s open room.<\/p>\n\n\n\n

Job Responsibilities:<\/strong><\/p>\n\n\n\n

    \n
  1. Build generalizable embodied AI foundation models, iterating and improving upon the training pipeline that has already been established within the team.<\/li>\n\n\n\n
  2. Abstract research problems from the above processes and conduct algorithm research, such as in latent action, VLA, world model, and RL.<\/li>\n<\/ol>\n\n\n\n

    Qualifications:<\/strong><\/p>\n\n\n\n

      \n
    1. PhD in Computer Science, Software Engineering, Electrical Engineering, or other related fields. Undergraduate and Master\u2019s students aspiring to scientific research are also welcome to apply.<\/li>\n\n\n\n
    2. Strong programming skills, good communication abilities, and a spirit of teamwork.<\/li>\n\n\n\n
    3. Basic knowledge in embodied intelligence\/deep learning, with the ability to understand top international conference papers in the field of embodied intelligence, and hands-on experience in deep learning.<\/li>\n\n\n\n
    4. Bonus points: Achievements in mathematics\/physics\/informatics competitions; experience in developing and researching embodied intelligence algorithms; publications in top international conferences and journals on embodied intelligence.<\/li>\n<\/ol>\n\n\n\n

      Internship Duration Requirements:<\/strong><\/p>\n\n\n\n

      Must obtain permission from your academic advisor and commit to at least three months of internship.<\/p>\n\n\n\n

      Please be sure to download and complete the application form (Application form link: https:\/\/aka.ms\/InternApplication (opens in new tab)<\/span><\/a>) and send it along with a complete English and Chinese resume (in PDF\/Word format) to: MSRAih@microsoft.com<\/a> & lizo@microsoft.com<\/a>. Please include “Research Intern for Machine Learning Group – Embodied AI research ” in the email subject line.<\/p>\n\n\n\n

       <\/p>\n\n\n\n

       <\/p>\n\n\n\n

      \u5c97\u4f4d\u540d\u79f0\uff1a<\/strong>\u673a\u5668\u5b66\u4e60\u7ec4—\u5177\u8eab\u667a\u80fd\u7814\u7a76\u5b9e\u4e60\u751f
      \u5de5\u4f5c\u6027\u8d28\uff1a<\/strong>\u5168\u804c\u5b9e\u4e60
      \u62db\u8058\u4eba\u6570\uff1a<\/strong>2-4\u4eba
      \u5de5\u4f5c\u5730\u70b9\uff1a<\/strong>\u5317\u4eac<\/p>\n\n\n\n

      \u7ec4\u522b\u7b80\u4ecb\uff1a<\/strong><\/p>\n\n\n\n

      \u6211\u4eec\u5173\u6ce8Embodied AI foundation model\u7684\u7814\u7a76\u3002\u5728\u7b97\u6cd5\u7814\u7a76\u65b9\u9762\uff0c\u6211\u4eec\u7684\u7814\u7a76\u5174\u8da3\u5305\u62ec\u4f46\u4e0d\u9650\u4e8elatent action learning\uff0cVLA\uff0c World Model\uff0cRL\u7b49\u7814\u7a76\u65b9\u5411\u3002\u6211\u4eec\u4e13\u6ce8\u4e8e\u4eceinternet-scale video data\u4e2d\u5b66\u4e60\u51b3\u7b56\u77e5\u8bc6\uff0c\u6765\u63d0\u9ad8embodied AI foundation model\u7684\u6cdb\u5316\u80fd\u529b\u3002\u5728\u8fc7\u53bb\u7684\u51e0\u5e74\u91cc\uff0c\u6211\u4eec\u5728\u9876\u7ea7\u56fd\u9645\u4f1a\u8bae\u548c\u671f\u520a\u4e0a\u53d1\u8868\u4e86\u591a\u7bc7\u8bba\u6587\u3002\u6211\u4eec\u63d0\u51fa\u7684IGOR\u6846\u67b6\uff0c\u901a\u8fc7\u4e92\u8054\u7f51\u89c4\u6a21\u7684\u89c6\u9891\u5b66\u4e60latent action, \u9996\u6b21\u5b9e\u73b0\u4e86\u4ec5\u901a\u8fc7latent action\u5c06\u4eba\u7c7b\u7684\u52a8\u4f5c\u8fc1\u79fb\u5230\u673a\u5668\u4eba\u4e0a\uff0c\u5f00\u542f\u4e86\u4eba\u7c7b\u5230\u673a\u5668\u4eba\u77e5\u8bc6\u8fc1\u79fb\u548c\u63a7\u5236\u7684\u65b0\u7684\u53ef\u80fd\u6027\u3002\u6211\u4eec\u5728RL\u4e5f\u4e0a\u6709\u6df1\u539a\u7684\u79ef\u7d2f\uff0c\u6211\u4eec\u7684\u7814\u53d1\u7684\u9ebb\u5c06 AI \u7cfb\u7edf Suphx \u6210\u4e3a\u9996\u4e2a\u5728\u56fd\u9645\u77e5\u540d\u4e13\u4e1a\u9ebb\u5c06\u5e73\u53f0 \u201c\u5929\u51e4\u201d\u4e0a\u8363\u5347\u5341\u6bb5\u7684 AI \u7cfb\u7edf\uff0c\u5176\u5b9e\u529b\u8d85\u8d8a\u8be5\u5e73\u53f0\u516c\u5f00\u623f\u95f4\u9876\u7ea7\u4eba\u7c7b\u9009\u624b\u7684\u5e73\u5747\u6c34\u5e73\u3002<\/p>\n\n\n\n

      \u5de5\u4f5c\u804c\u8d23\uff1a<\/strong><\/p>\n\n\n\n

        \n
      1. Build\u6cdb\u5316\u7684embodied AI foundation model\uff0c\u5728\u7ec4\u5185\u5df2\u7ecf\u642d\u5efa\u597d\u7684training pipeline\u4e0a\u505a\u8fed\u4ee3\u7684\u6539\u8fdb\u548c\u63d0\u9ad8<\/li>\n\n\n\n
      2. \u4ece\u4e0a\u8ff0\u8fc7\u7a0b\u4e2d\u62bd\u8c61\u7814\u7a76\u95ee\u9898\uff0c\u8fdb\u884c\u7b97\u6cd5\u7814\u7a76\uff0c\u6bd4\u5982\u5728latent action\uff0cVLA\uff0cworld model\uff0cRL\u65b9\u9762\u7684\u7814\u7a76<\/li>\n<\/ol>\n\n\n\n

        \u4efb\u804c\u8981\u6c42\uff1a<\/strong><\/p>\n\n\n\n

          \n
        1. \u8ba1\u7b97\u673a\u79d1\u5b66\u3001\u8f6f\u4ef6\u5de5\u7a0b\u3001\u7535\u5b50\u5de5\u7a0b\u6216\u5176\u5b83\u76f8\u5173\u4e13\u4e1a\u535a\u58eb\uff0c\u6709\u5fd7\u4e8e\u79d1\u7814\u7684\u672c\u79d1\u751f\u6216\u7855\u58eb\u4e5f\u6b22\u8fce\u7533\u8bf7\uff1b<\/li>\n\n\n\n
        2. \u5177\u6709\u8f83\u5f3a\u7684\u7f16\u7a0b\u5b9e\u73b0\u80fd\u529b\uff0c\u826f\u597d\u7684\u6c9f\u901a\u80fd\u529b\u548c\u56e2\u961f\u534f\u4f5c\u7cbe\u795e<\/li>\n\n\n\n
        3. \u5177\u5907\u4e00\u5b9a\u7684\u5177\u8eab\u667a\u80fd\/\u6df1\u5ea6\u5b66\u4e60\u57fa\u7840\uff0c\u80fd\u591f\u8bfb\u61c2\u9876\u7ea7\u56fd\u9645\u4f1a\u8bae\u5177\u8eab\u667a\u80fd\u65b9\u9762\u7684\u8bba\u6587\uff0c\u6709\u6df1\u5ea6\u5b66\u4e60\u4e0a\u624b\u7ecf\u9a8c<\/li>\n\n\n\n
        4. \u52a0\u5206\u9879\uff1a\u6709\u6570\u5b66\/\u7269\u7406\/\u4fe1\u606f\u5b66\u7ade\u8d5b\u65b9\u9762\u7684\u6210\u7ee9\uff1b\u6709\u5177\u8eab\u667a\u80fd\u7b97\u6cd5\u7684\u5f00\u53d1\u548c\u7814\u7a76\u7ecf\u5386\uff1b\u5728\u9876\u7ea7\u56fd\u9645\u4f1a\u8bae\u548c\u671f\u520a\u4e0a\u53d1\u8868\u8fc7\u5177\u8eab\u667a\u80fd\u65b9\u9762\u7684\u5de5\u4f5c\u3002<\/li>\n<\/ol>\n\n\n\n

          \u5de5\u4f5c\u65f6\u95f4\u8981\u6c42\uff1a<\/strong><\/p>\n\n\n\n

          \u80fd\u83b7\u5f97\u5bfc\u5e08\u8bb8\u53ef\u5e76\u4fdd\u8bc1\u81f3\u5c11\u4e09\u4e2a\u6708\u7684\u5b9e\u4e60\u3002<\/p>\n\n\n\n

          \u8bf7\u52a1\u5fc5\u4e0b\u8f7d\u5e76\u586b\u5199\u7533\u8bf7\u8868\uff08\u7533\u8bf7\u8868\u94fe\u63a5\uff1ahttps:\/\/aka.ms\/InternApplication (opens in new tab)<\/span><\/a>\uff09\u5e76\u5c06\u5176\u4e0e\u5b8c\u6574\u7684\u4e2d\u82f1\u6587\u7b80\u5386\uff08PDF\/Word\u5f62\u5f0f\uff09\u4e00\u540c\u53d1\u9001\u81f3\uff1a MSRAih@microsoft.com<\/a> & lizo@microsoft.com\uff0c\u90ae\u4ef6\u6807\u9898\u4e2d\u6ce8\u660e\uff1a\u673a\u5668\u5b66\u4e60\u7ec4—\u5177\u8eab\u667a\u80fd\u7814\u7a76\u5b9e\u4e60\u751f<\/strong>\u3002<\/p>\n\n\n\n

          \u4e86\u89e3\u66f4\u591a\u201c\u660e\u65e5\u4e4b\u661f\u5b9e\u4e60\u751f\u8ba1\u5212\u201d\uff0c\u8bf7\u8bbf\u95ee\uff1a\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u201c\u660e\u65e5\u4e4b\u661f\u201d\u5b9e\u4e60\u751f\u9879\u76ee – Microsoft Research<\/a><\/p>\n","protected":false},"featured_media":0,"parent":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","msr_opportunity_details":{"preferred_date":"","date_type":"preferred_date","featured":"","hide_legal_copy":""},"msr_opportunity_pubdate":"2025-07-03","msr_hidden_opportunity":false,"msr_opportunity_poc":{"name":"","email":"","phone":""},"msr_opportunity_hta":"","msr_opportunity_external_link":"no","msr_opportunity_legal_copy":"","msr-author-ordering":[],"footnotes":""},"research-area":[13556],"msr-region":[197903,256048],"msr-job-opportunity-type":[234821],"msr-city":[243628],"msr-locale":[268875],"msr-program-audience":[],"msr-impact-theme":[],"class_list":["post-1143797","msr-job-opportunity","type-msr-job-opportunity","status-publish","hentry","msr-research-area-artificial-intelligence","msr-region-asia-pacific","msr-region-global","msr-job-opportunity-type-internship","msr-city-beijing-china","msr-locale-en_us"],"msr_opportunity_cta":"","msr_opportunity_subdate":"","msr_opportunity_subdate_type":"preferred","msr_opportunity_pubdate":"2025-07-03","msr_opportunity_legal_copy":"","msr_opportunity_poc":{"name":"","email":"","phone":""},"msr_opportunity_type":["Internship"],"msr_research_lab":[199560,1012650],"related-researchers":[],"msr_impact_theme":[],"related-groups":[269241],"related-projects":[],"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-job-opportunity\/1143797","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-job-opportunity"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-job-opportunity"}],"version-history":[{"count":5,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-job-opportunity\/1143797\/revisions"}],"predecessor-version":[{"id":1167279,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-job-opportunity\/1143797\/revisions\/1167279"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=1143797"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=1143797"},{"taxonomy":"msr-region","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-region?post=1143797"},{"taxonomy":"msr-job-opportunity-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-job-opportunity-type?post=1143797"},{"taxonomy":"msr-city","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-city?post=1143797"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=1143797"},{"taxonomy":"msr-program-audience","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-program-audience?post=1143797"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=1143797"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}