Headline

Applied Scientist 2 @ Microsoft | Recommendation & NLP | Focused on foundational ML work

About

I build ML systems where the solution becomes the foundation for what comes next. Right now at Microsoft, I work on Ads Trust & Safety — owning the end-to-end ML for a core ad policy: detection, measurement, and the infrastructure that determines whether enforcement is actually working across Search and Native platforms. Before moving into ATS, I spent a year and a half building retrieval models for ad selection: intent-based encoder models, evaluation frameworks, and the data pipelines behind it all. Before Microsoft, I shipped AI systems to millions of users at Zomato everything from image quality scoring, Personalized Photo Cakes to PoC for automated ad creation. At Enterpret, I researched text similarity at scale, spending most of my time on data preparation because that's usually where projects succeed or fail. What drives me: I care about foundational work. The kind where if you get it right, it unlocks progress for everyone who comes after, be it data, modeling or evaluations!

Location

Bengaluru

AI stack

hugging face

mcp

huggingface

ChatGPT

Work experience

Applied Scientist 2 at Microsoft AI, Mar 2026 - Present · 7 mos, Bengaluru

Applied Scientist at Microsoft AI, Sep 2024 - Feb 2026 · 1 yr 6 mos, On-site, - Developing next-gen retrieval models for ad selection across Microsoft Audience Network - Built unified evaluation framework adopted as standard across regional ML teams - Established foundational data quality practices now used across MSAN globally - Driving production impact on ad monetization through principled ML approaches

Machine Learning Engineer at Zomato, Jan 2024 - Aug 2024 · 8 mos, Gurugram, - Built image quality system evaluating every food image shown to millions of users - Led the PoC for automated ad creation using generative models for brand-specific content - Launched photo cake feature—personalized orders at scale across India

Machine Learning at Enterpret, Jun 2023 - Dec 2023 · 7 mos, On-site, - Researched text similarity methods for business intelligence at million-text scale - Created data frameworks and supervised annotation teams for quality training data - Combined prompt engineering, fine-tuning, and novel loss functions for strong results

Machine Learning Intern at Department of Computer Science and Engineering, IIT Jodhpur, Jul 2021 - May 2022 · 11 mos, India, Guide: Dr.Santanu Chaudhary - My main work includes searching and developing ML models for the purpose of Document Layout Understanding - To be more specific, the ML models are inspired by the transformers network, and there are some novel losses that have been incorporated for the same.

Education

Bachelor of Technology - B Tech, Electrical Engineering, Indian Institute of Technology Jodhpur

Public links

github: https://github.com/uakarsh

youtube.com: https://youtube.com/@akarshupadhyay1241

uakarsh.github.io: https://uakarsh.github.io/index.html