About this role

<b>Overview</b><br><div><p><span>Microsoft Teams</span><span> is the hub for teamwork that integrates all the people, content, and tools your team needs to be more engaged and effective. It is core to Microsoft’s modern work, modern life & modern education value prop. </span><span>We are reinventing the way people communicate and work together across the globe.  </span><span> </span></p></div><div><p><span>We are looking to hire a PhD (or published MSc) candidate for a part-time internship to join </span><span>CMD Labs</span><span> – an applied science team within Microsoft Teams – to work on improving collaborative agents by applying existing or novel research. </span><span> </span></p></div><div><p><span>Our flagship AI application for Teams is the Copilot agent that works in multi-party collaborative settings (meetings, group chats, channels) and enables groups of humans and agents to more efficiently achieve their shared goals. While most human/agent interactions are textual, within live Teams meetings the agent also interacts in real time via speech. Unlike personal agents (both audio and text), multi-party collaborative agents are faced with significant challenges in implicit context understanding, proactivity, and conversational capabilities.</span><span> </span></p></div><div><p><span>The intern will be onboarded to our current research directions, and to our non-confidential datasets and related code, and work on improving the existing algorithms as well as proposing novel solutions, in any of the following topics:</span><span> </span></p></div><div><ul style="list-style-type: disc;"><li><p><span>Training/fine-tuning </span><span>speech-enabled models</span><span> (e.g. Qwen-omni-8) to support speech in multi-party settings such as business meetings.</span><span> </span></p></li></ul></div><div><ul style="list-style-type: disc;"><li><p><span>Improve and devise new methodologies for </span><span>evaluation </span><span>of audio/text collaborative agents.</span><span> </span></p></li></ul></div><div><ul style="list-style-type: disc;"><li><p><span>Improve and devise better </span><span>RL recipes</span><span> for training/fine-tuning collaborative models including synthetic data generation, and the development and calibration of graders.</span><span> </span></p></li></ul></div><div><p><span>Given substantial scientific novelty of the approach and results, collaboration on a mutual publication is encouraged.</span><span> </span></p></div><br><br><b>Responsibilities</b><br><div><div><ul style="list-style-type: disc;"><li><p><span>Conduct experiments, create and validate metrics, and develop candidate algorithms to improve the existing capabilities of current models and agent implementations in any of the above topics.</span><span> </span></p></li></ul></div><div><ul style="list-style-type: disc;"><li><p><span>Collaborate closely with CMD Labs researchers and engineers to leverage existing assets, datasets, and ensure results can contribute to the product.</span><span> </span></p></li></ul></div></div><div><div><ul style="list-style-type: disc;"><li><p><span>Embody our </span><a href="https://nam06.safelinks.protection.outlook.com/?url=https%3A%2F%2Fcareers.microsoft.com%2Fus%2Fen%2Fculture&data=05%7C01%7Cyashj%40microsoft.com%7C45b5b9f502f0431544c808db4692dd76%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C638181368816489204%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=7oKBX7BgaxfFOI8A4y445FP1VlXXydU9bCsMLZSd0t8%3D&reserved=0" target="_blank" rel="noreferrer noopener"><span>culture</span></a><span> and </span><a href="https://nam06.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.microsoft.com%2Fen-us%2Fabout%2Fcorporate-values&data=05%7C01%7Cyashj%40microsoft.com%7C45b5b9f502f0431544c808db4692dd76%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C638181368816489204%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=9NVcjeaFP9oTdAgsGZXG1OIWRoTacWw1SjYD7j4uoFM%3D&reserved=0" target="_blank" rel="noreferrer noopener"><span>values</span></a><span>.</span><span> </span></p></li></ul></div></div><br><br><b>Qualifications</b><br><div><p><span>Required</span><span> </span></p></div><div><ul style="list-style-type: disc;"><li><p><span>Currently enrolled in a PhD program (or published candidate in MSc program) in Computer Science, Electrical or Computer Engineering, Statistics, or a related field.</span><span> </span></p></li></ul></div><div><ul style="list-style-type: disc;"><li><p><span>Practical experience in training, fine-tuning, and prompt engineering of transformer models or LLMs.</span><span> </span></p></li></ul></div><div><ul style="list-style-type: disc;"><li><p><span>Practical Python coding experience leveraging modern coding agents.</span><span> </span></p></li></ul></div><div><p><span>Preferred</span><span>    </span><span> </span></p></div><div><ul style="list-style-type: disc;"><li><p><span>Field of research and publications directly related to Audio LLMs or collaborative models and agents.</span><span> </span></p></li></ul></div> <br><p>This position will be open for a minimum of 5 days, with applications accepted on an ongoing basis until the position is filled.</p><br><hr><br><p>Microsoft is an equal opportunity employer. All qualified applicants will receive consideration for employment without regard to age, ancestry, citizenship, color, family or medical care leave, gender identity or expression, genetic information, immigration status, marital status, medical condition, national origin, physical or mental disability, political affiliation, protected veteran or military status, race, ethnicity, religion, sex (including pregnancy), sexual orientation, or any other characteristic protected by applicable local laws, regulations and ordinances. If you need assistance with religious accommodations and/or a reasonable accommodation due to a disability during the application process, read more about <a href="https://careers.microsoft.com/v2/global/en/accessibility.html"><b><u>requesting accommodations.</u></b></a></p>