Suresh Parthasarathy

À propos

I am a research developer with the Applied Sciences and Engineering group at MSR India. I am broadly interested in the areas of machine learning and applied ML projects. My current focus is to combine program synthesis and analysis algorithms with ML models. Here are some of my current projects:

• Jigsaw: In this project, we combine large language models like Codex and GPT-3 with program analysis and synthesis techniques to ensure code correctness. We propose one such instantiation with respect to Python Pandas in our ICSE’22 paper (opens in new tab). Our current focus is on JavaScript.
• HDEF: In this project, we proposed a novel disjunctive synthesis algorithm which combines noisy labels from a ML model with program synthesis to automatically synthesize extraction wrappers. Here is our PLDI’19 paper (opens in new tab) on the same. This work is also deployed in Microsoft Exchange.
• Omega (Combining ML model and rules): This is a more exploratory project where we seek to come up with a framework where ML models and domain rules (human-written or synthesized) can co-exist and evolve with time.
• Ads decorations: In this project, I work closely with the Bing Ads team and lead an effort on entity extractions from Ad landing pages to make an Ad snippet more useful. This effort has lead to many mainstreaming in Bing Ads,  in multiple markets, resulting in significant revenue gains.
• Azure Purview: Here, I lead an effort to introduce AI-based solutions to various data governance tasks. The main challenges in this space pertain to data compliance and scale.

I’ve had the pleasure of mentoring a few amazing Research Fellows (opens in new tab) and interns:

• Arnav Chavan (opens in new tab), working with me on HDEF project
• Atharv Sonwane (opens in new tab), working with me on Jigsaw project
• Mohammad Taufeeque (opens in new tab) (co-mentor), worked on Omega. Joining Georgia Tech for his PhD.
• Naman Jain (opens in new tab) (co-mentor), worked on Jigsaw. Currently pursuing his PhD at UC Berkeley.
• Skanda Vaidyanath (opens in new tab) (co-mentor), worked on Jigsaw. Currently pursuing his Master’s at Stanford
• Lincy Pattanaik (opens in new tab), worked on Robust Synthesis project. Currently pursuing her Master’s at UMass.
• Anirudh Khatry (opens in new tab), worked on Robust Synthesis project, joined Microsoft PROSE team as Research Fellow.
• Priyan Vaithilingam (opens in new tab), contributed to InLook and CodeTalk projects. Currently pursuing his PhD at Harvard
• Venkatesh Potluri (opens in new tab), contributed to CodeTalk project. Currently pursuing his PhD at UW
• Shruti Rijhwani (opens in new tab), contributed to InLook and Fuzzy Search projects. Currently pursuing her PhD at CMU
• Ishani Ghose (opens in new tab), contributed to HDEF project. Currently pursuing masters at Rutgers.
• Abhinav Duggal (opens in new tab), contributed to InLook project. Joined Microsoft Bing Ads, Bangalore
• Chetan Bansal (opens in new tab), contributed to Delve project. Joined Microsoft Research
• Vishwajit Kolathur (opens in new tab), contributed to Delve project. Joined Amazon AWS, Seattle
• Pranav Ramarao (opens in new tab), contributed to InLook and CodeTalk projects. Joined Google, MV
• Afroz Ahamad (opens in new tab), contributed to CodeTalk project. Joined Google, Bangalore
• and many more…

I have worked on ML projects in Delve/Office graph, resource-constrained environments, email relevance and reviews ranking to name a few. I have  contributed to the ELL (opens in new tab) and EdgeML (opens in new tab) libraries. Many ideas from my hackathon project InLook (opens in new tab) have found their way into Outlook. I also led an effort in making programming accessible to visual impaired developers. My project CodeTalk (opens in new tab) won a honorable mention in CHI’18 conference.

Before coming to MSR India, I did a short one year stint at Yahoo Research, Bangalore where I worked with Dr. Rajeev Rastogi (opens in new tab) on problems related to product title de-duplication, along with other problems on entity matching.

I was a software developer in Bing team at Microsoft India Development Center for three years before this, wherein I was responsible for Bing instance answers on events and tennis related queries, Bing events and attraction verticals and problems related to data de-duplication. Specifically, I worked on Bing-FanSnap integration (opens in new tab) which garnered positive press.

Prior to this, I completed my Masters in Computer Science from IIT Bombay (opens in new tab)under Prof. S. Sudarshan (opens in new tab).