S

Shikhar Sharma

AI Researcher

India3 yrs 5 mos experience
AI EnabledAI ML Practitioner

Key Highlights

  • Over a decade of deep learning expertise
  • Led machine learning teams at Microsoft
  • Innovative contributions to AI and NLP projects
Stackforce AI infers this person is a SaaS expert with a strong focus on AI and machine learning solutions.

Contact

Skills

Core Skills

Machine LearningDeep LearningNatural Language Processing (nlp)

Other Skills

AIAlgorithmsArtificial Intelligence (AI)DockerImage SynthesisMicrosoft AzureMicrosoft Azure Machine LearningPrecipitation NowcastingPyTorchPythonRecurrent Neural NetworksRoboticsTeam LeadershipTensorBoardWeather Forecasting

About

As an engineer with over a decade of deep learning expertise, I've spearheaded machine learning teams towards the successful completion of groundbreaking projects. My forte lies in developing innovative tools that harness machine learning and artificial intelligence to boost inspiration, productivity, and capabilities. I have a strong research and engineering background with an M.Sc. from University of Toronto, Canada and a B.Tech. from IIT Kanpur, India. During the course of my career, I have also had the opportunity to lead a team of exceptional machine learning engineers and supervise a number of outstanding interns.

Experience

Microsoft

3 roles

Principal Applied Scientist

Promoted

Sep 2024Present · 1 yr 6 mos

  • [Bing Multimedia] I built the first LLM-based online evaluation metric for Bing's Image Feed. I was then made Tech Lead for the Ranking team consisting of 6 ML engineers. Under my leadership, we migrated the ranking stack from traditional ML models to deep learning-based ones. We deploy and maintain rankers for both the Image Feed and Related Content products, commanding a Daily Active User (DAU) count of millions of users.
Deep LearningMachine LearningArtificial Intelligence (AI)PythonPyTorch

Senior Research SDE

Promoted

Sep 2019Aug 2024 · 4 yrs 11 mos

  • [Turing] I worked across several Microsoft products leveraging AI. The highlight was my work on Sydney (a.k.a. Microsoft Copilot)'s defensive classifier, making it capable of handling conversational context leading to a 10% increase in conversation length, 27% drop in offensiveness, 50% drop in compute requirement. I also worked on long-horizon projects such as high-quality image synthesis for user search queries.
  • [Weather Forecasting] I was one of the early deep learning advisors and contributors to the precipitation nowcasting product which is now deployed at Microsoft Weather and boasts millions of users.
Natural Language Processing (NLP)Machine LearningDeep LearningAI

Research SDE II

Feb 2017Aug 2019 · 2 yrs 6 mos

  • [Microsoft Research] My research focused on image synthesis using captions, conversations, and spatial layouts. I expanded my prior dialogue research to multi-modal dialogue systems. My research agenda at MSR led to several publications at top conferences. Part of my time was devoted to integrating state-of-the-art research into Microsoft products among which the most prominent partnership was with Microsoft Support on their query understanding system.
Natural Language Processing (NLP)Machine LearningImage Synthesis

Maluuba

Research Scientist

Mar 2016Feb 2017 · 11 mos · Montreal, Quebec, Canada

  • At Maluuba I worked with the dialogue group on task-oriented dialogue systems. My work resulted in 2 patents and 4 publications. Part of my work focused on improving natural language generation using recurrent neural networks and utilizing unsupervised metrics for their evaluation. The Maluuba/nlg-eval GitHub repository for which I was the main contributor boasts 1000+ stars and 100+ forks. I also worked on the collection and release of the Frames dataset, which was the largest multi-frame task-oriented dialogue dataset at the time. Subsequently, I worked on end-to-end trainable models for it. Maluuba was acquired by Microsoft in 2017.
Natural Language Processing (NLP)Machine LearningRecurrent Neural Networks

Cornell university

Research Intern

May 2013Jul 2013 · 2 mos · Tompkins County, New York, United States

  • I worked with Prof. Ashutosh Saxena and Ashesh Jain on learning preferences over trajectories on robots such as the Baxter. Our approach required a non-expert user for training and the preferences we learned were governed by objects and human interactions in the environment. Our work was featured on TechCrunch, Cornell CS News, NBC, KurzweilAI and several other news websites garnering over 140,000 YouTube views.
Machine LearningRobotics

Education

University of Toronto

Master of Science (M.Sc.) — Computer Science

Jan 2014Feb 2016

Indian Institute of Technology, Kanpur

Bachelor of Technology (B.Tech.) — Computer Science & Engineering

Jan 2010Jan 2014

Stackforce found 100+ more professionals with Machine Learning & Deep Learning

Explore similar profiles based on matching skills and experience