PROSE Research Fellowship Program

Scale up impact while nurturing globally diverse talent and growing research leaders

 

I am a product of a typical graduate school education that focuses on individual excellence, and it took me the first half of my research career to realize that we are in an era where big scientific discoveries are unlikely to come from a lone genius working in isolation. Today’s disruptive science is a process of layering, combining ideas, and collaboration involving big teams—of individuals that bring different skills to the table, focus, and trust each other to work together. However, such structures are hard to create due to talent shortage and funding limitations as I have experienced over the second half of my career. 

It has been 5 years since we formed the PROSE team to advance the state-of-the-art in Program Synthesis and incorporate those innovations inside Microsoft products. The result has been delightful experiences for a huge customer base (with a million+ invocations/week of the Flash Fill feature in Excel, a million+ accepted suggestions/month of the Intellicode Suggestions feature in VS, and a million+ invocations/month of the data connectors in PQ) and a multitude of papers in top-tier conferences across various research areas (several winning best paper awards and test-of-time most-influential paper awards).

– Sumit Gulwani, upon founding the PROSE Research Fellowship program in late 2020

 

The key to the PROSE team’s success has come through a 4-step process which determines our culture: 

1. The most important part of solving a problem is the definition, so we focus on: 
   • Customer connection – We are very savvy in our choice of problems, ensuring they are derived from real customer scenarios. 
   • Impact, not cleverness – Rather than finding the hardest problem, we choose the problem with the most impact.  
2. The second step is the solution.  To drive solutions, we use: 
   • Cross-disciplinary innovation – We thrive on leveraging the diverse expertise of our researchers across software engineering, programming languages, formal methods, HCI, ML. 
   • Foundational framework-based approach – We build generalized frameworks (both in our research approach and our engineering deliverables) instead of creating point solutions, which allows us to have broad impact with a lean team. 
3. We deliver our solutions through: 
   • Blended research & engineering – We create shipping-quality implementations of our algorithms that do not need rework to be used in Microsoft products. 
   • Trusted relationships – We partner with product teams to integrate our APIs inside product code bases through deep, trusted relationships built on a track-record of mutual success. 
4. Finally, we relentlessly engage in continuous improvement of both: 
   • What we build – We capitalize on our trusted partner relationships to improve our understanding of customer pain points and then cycle back to the first step and start the process again. 
   • How we build it – We always look for ways to increase our effectiveness. This Research Fellowship program is a key example; it fundamentally alters the playing field. 

The Research Fellowship Program has three goals that feed into each other, and are related to our culture and process: 

– We are growing a new class of junior researchers with a value system that is built around our culture and process. These serve in contrast to the typical graduate education focus on individualistic values and solving novel abstract problems. We believe our process can equip these researchers to increase their impact on academia, industry, and our customers.  

– At the same time, we are increasing the impact of our research team by creating opportunities for researchers to manage junior level but high-quality talent and to serve as ambassadors for our culture and process. 

– Through these efforts we increase the available talent pool of both young researchers who may not have access to good graduate education and of experienced, post-PhD researchers who might otherwise choose academia over industry to have a team of young talent help them pursue their research. 

 

What?

A global Research Fellowship program inspired by other research fellowship and residency programs around the company but continuously evolving based on what we learn as we go. 

• We hire and train top undergraduates or master’s students from top Universities globally to work from their local region for 1-2 years of temporary appointment with a path to full-time employment and the opportunity of a career as a researcher. 
• Research Fellows start with a 1-year appointment renewable to 2-years including both training and work as part of existing blended research and engineering projects.   

• Research fellows who demonstrate sufficient research and engineering skills (typically after 1-2 years) become eligible either for an offer of full-time employment or for our support of their application to a top PhD program (in the form of strong recommendation letters). 
• When there is the combination of demonstrated skills, mutual interest, and funding availability, we make offers for research fellows to join as an “associate researcher” at L59 or L60.  
• Associate researchers who perform well are eventually promoted to “associate researcher 2” at L61 and then to level L62.  This process from initially joining Microsoft full-time to becoming ready for L62 typically takes about 4 years. 
• When they are ready for L62, an independent committee decides whether we put them on a “researcher” career path (without a PhD but based on their research track record) or onto an engineering career path. 
• This process is a give-and-take relationship. Researchers gain the ability to delegate tasks and expand the pool of effort dedicated to solving the problems they identify. As research fellows demonstrate success, they command more research-intensive work items. Over a journey of around 6 years, we invest in junior talent and grow a cadre of research leaders who are well-positioned to have high impact anywhere at Microsoft.   

Why this approach?

• While the opportunities for good K-16 education are spread throughout the globe, the opportunities for good graduate education are concentrated in small parts of the globe.  This leads to massive under-utilization of talent that could be in the business of science. 
• The top-tier talent in India and other places where good graduate education is unavailable locally generally goes to top PhD Universities in the US and Europe or opts to pursue non-research paths such as entrepreneurship.  We attract part of this talent to build their careers, stay at Microsoft in their home countries, and build up their local research communities. 
• We frequently have discussions with interested students, often with strong software engineering skills, who desire to be more involved in research and innovation but are unable to pursue a full-time PhD. They may not be able to relocate for an extended period. Or they want to keep both engineering and research career paths open and are not interested in the large commitment of a PhD. This offering is uniquely appealing to these candidates.   
• In addition to expanding the opportunities for students and to increasing Microsoft’s pool of early in career researchers, this also increases the appeal to post-PhD researchers who may otherwise have considered careers in Academia because of the opportunity for expanded impact that comes from leading a team of students.  In industry, individual researchers often have less opportunity to work on larger problems because of limitations in the number of individuals they can direct in their research.  Allowing them to manage research fellows increases their impact and addresses this concern when they compare career options. 

How?

• Researchers provide direction to Research Fellows thus reducing ambiguity in problem definition and strategy while creating space for exploration, experimentation, prototyping and participating in paper writing.  All of this is done in the context of real-world, customer-centric projects and blended research and engineering.  Thus, Research Fellows get a unique opportunity to learn while amplifying the output of researchers on the team. 
• The team is fully invested in the growth of this junior level talent. Research Fellows budget 20% of their time for training, around half of which uses curated external resources and around half is internal material created by the PROSE team (including research background, engineering skills, and storytelling). We call this training the PROSE Academy.   
• Associate researchers continue to participate in the PROSE Academy but progressively have more independent research responsibilities and a greater role in product delivery and paper writing. 
• The program is intentionally cooperative.  Research fellows and associate researchers support one another and those with more experience mentor newer members of the program. 

• We create visibility in the academic community to identify and recruit good candidates by giving keynotes at computer science conferences, giving guest lectures at universities, and collaborating with professors.  

Supporting Work Across Time Zones 

We are keenly aware that global partnerships require addressing time-zone challenges, especially with more junior talent. We apply learnings from our successful MSRI collaborations, our experience with team members in other countries, and best practices in the field. For example: 

• We adopt a remote-first approach to meetings including always using Teams, recording meetings, sharing video stories and demos, increasing investment in documentation and other best practices in asynchronous communication. The importance of these practices is cemented by the fact that multiple members of team leadership work remotely either primarily or entirely. 

• Managers of research fellows are sensitized to the importance of being available to and regularly connecting.  They establish mutually acceptable schedules for sync-ups and ad hoc discussions which may require them to be available early or late to appropriately support their research fellows. 
• We create pods with at least one researcher and 2 research fellows working on the same project. This allows the research fellows to support each other and simplifies coordination with a smaller group rather than the whole team. 
• Cross-team meetings are scheduled only at times that work across time zones. 
• Team members are increasingly geographically diverse including some on the North American east-coast and in Europe which reduces time zone differences to other global geographies. 
• Where possible we seek out local mentors to expand research fellows’ network on-the-ground.  

Success Criteria 

• Retention of research fellows and progression through the program to full researchers. 
• Product impact directly by research fellows or indirectly by enabling their researchers. 
• Papers published by or with substantial contributions from the research fellows. 

Frequently Asked Questions 

• What is the interview process like? What references should I look at to best prepare for the research interviews?
  Interviewing will consist of research discussions with few researchers on the team and an engineering round. You should be able to describe your own prior research work clearly. You can also ask us questions about our research work during the interviews.  The engineering interview will be led by engineers on our team. More than competitive coding, we look for ability to (1) think on the feet, (2) write good and well-documented code, (2) design tests, and (4) debug code effectively. Coding is a tool to have impact, but not the end goal.

• Is this a paid position? How do the pay and benefits compare with the MSR India RF program?
  Yes, this is a paid position. The pay and benefits are the same as the MSR India RF positions.

• Is a semester-long (4-6 months) internship possible through the program?  
  No. We are investing in people as much as in projects. This has high startup cost, and we also have an additional challenge of remote work. Hence, we are looking for at least 1 year appointments for this to be meaningful.

• How does it work across time zones?
  We adopt a remote-first approach in the team, recording meetings, sharing video stories and demos, increasing investment in documentation and other best practices in asynchronous communication. The importance of these practices is cemented by the fact that multiple members of team leadership work remotely either primarily or entirely.  Managers of research fellows are sensitized to the importance of being available to and regularly connecting.  They establish mutually acceptable schedules for sync-ups and ad hoc discussions which may require them to be available early or late to appropriately support their research fellows.

• Will this fellowship lead to a full-time position in India or the United States?
  While the fellowship is a temporary position based out of India, we have previously made full-time offers to RFs for positions in India. However, this is contingent on strong and sustained performance on the part of the fellow, and on the availability of positions. The full-time offers are for associate researcher positions and we will aspire to grow the fellow into strong independent researchers over a period of 3-4 years.

• Is the PROSE team the one developing Copilot?
  Microsoft Copilot products are companywide efforts, with a myriad of teams being involved in various roles. Our team is involved in research that is crucial to various Copilot products including Microsoft Visual Studio Copilot, PowerQuery Copilot, and Microsoft Excel Copilot. See our publications page for a sample of our recent work that has contributed to the Copilot ecosystem.

• How do I apply for the program? When is the deadline to apply?
  To apply, send an email to [email protected]. Please send a CV and optionally a few paragraphs on your aspirations and why you are interested in this opportunity.  There’s no application deadline, we accept applications on a rolling basis. See more details in our career opportunity page.

• Is a semester-long (4-6 months) internship possible through the program?
  No. We are investing in people as much as in projects. This has high startup cost, and we also have an additional challenge of remote work. Hence, we will be looking for at least 1 year appointment for this to be meaningful, with preference given to candidates who aspire to engage with us on longer journeys.  

• How is it different from MSRI RF program? 
  After the predoc appointment, we aspire to support a new career pathway for interested candidates to continue a full-time, research-track appointment with our team, working out of Microsoft India (an experience that can be compared with that of enrolling into a Ph.D. program at a top University, but with different pros/cons). Our team is a hybrid of research and engineering. We aim to not only make research contributions (publications), but also have product impact. Our program provides opportunities to develop both types of skills, which can serve as a precursor to a career in either.  

• Should I apply to both the PROSE Research Fellowship and MSRI Research Fellowship programs? 
  Yes, you do need to apply to both programs. While the PROSE team’s research fellowship program is modelled on MSR India’s program, they are separate. We strongly suggest that you apply to both programs and maximize the chance of finding the right opportunity that will suit you.

• Do I get to go to Redmond, USA to work with the group now or in future? 
  Travel for business reasons has typically been supported by Microsoft, whether to present at a conference or for an in-person visit to collaborators. However, the appointment will primarily be based out of India. However, the candidates would be welcome to apply to any job posting inside Microsoft global if they would so desire, after they reach a certain tenure.  

• Will it be possible for me to pursue projects that might not align with the group? 
  The real fun in working with our group is to leverage the unreasonable advantage that we have, created by nurturing connections with certain product groups (to deliver impact and leverage customer feedback), and by building on top of the technologies that we have developed. There is always room for investing in newer technologies and to make new pitches to product leaderships for shipping impact through their products; and both are strongly encouraged! Within this scope, we expect the research fellow to both learn and innovate, to learn to deliver success as we know it, and to lead us to deliver success in new areas/ways. However, if the goal is to only write papers, a university environment would be better. If the goal is to create impact in areas where Microsoft doesn’t operate in, then a better option might be to join those other organizations that operate in those areas. 

---

This proposal is always a work in progress. Feedback and partnership are most welcome. For any feedback or questions, contact Sumit Gulwani at [email protected]

Opens in a new tab