Uploaded July 2026 | Updated September 2026, 2 weeks ago
Merge conflicts are an unavoidable part of collaborative software development, often slowing teams down and creating friction during the development process. Recent advances in Large Language Models (LLMs) have opened new possibilities for understanding code context and assisting developers in resolving conflicts more efficiently.
In this talk, Advitya Gemawat, ML Engineer at Microsoft, will explore how LLMs can be applied to merge conflict resolution, the challenges involved in understanding code changes across branches, and the opportunities for building intelligent developer tools. Drawing from his experience building machine learning and AI systems at Microsoft, Advitya will discuss practical approaches, current limitations, and the future of AI-assisted software engineering.
0:00 - Intro
3:49 - Merge Conflict: Problem Space
5:17 - Why vanilla LLMs don’t know how to fix conflicts
9:03 - Impact of merge conflicts in large software projects
10:36 - Supervised Fine-tuning use-case
12:44 - LLM Customization Insights
18:50 - Evaluation Metrics
20:54 - Structured Outputs
23:23 - “One thing I’d do differently in my Python code”
25:24 - Your Coding Agent is NOT going to tell you this!
28:10 - Navigating Service Limits with Token Estimation
Follow Advitya on the handles below:
LinkedIn linkedin.com/in/agemawat
YouTube youtube.com/@AdvityaGemawat
Instagram instagram.com/advitya_17
Facebook facebook.com/Advitya17
WhatsApp whatsapp.com/channel/0029VaBAHaY6buMPdgOn2R2w
Merge conflicts are an unavoidable part of collaborative software development, often slowing teams down and creating friction during the development process. Recent advances in Large Language Models (LLMs) have opened new possibilities for understanding code context and assisting developers in resolving conflicts more efficiently.
In this talk, Advitya Gemawat, ML Engineer at Microsoft, will explore how LLMs can be applied to merge conflict resolution, the challenges involved in understanding code changes across branches, and the opportunities for building intelligent developer tools. Drawing from his experience building machine learning and AI systems at Microsoft, Advitya will discuss practical approaches, current limitations, and the future of AI-assisted software engineering.
0:00 - Intro
3:49 - Merge Conflict: Problem Space
5:17 - Why vanilla LLMs don’t know how to fix conflicts
9:03 - Impact of merge conflicts in large software projects
10:36 - Supervised Fine-tuning use-case
12:44 - LLM Customization Insights
18:50 - Evaluation Metrics
20:54 - Structured Outputs
23:23 - “One thing I’d do differently in my Python code”
25:24 - Your Coding Agent is NOT going to tell you this!
28:10 - Navigating Service Limits with Token Estimation
Follow Advitya on the handles below:
LinkedIn linkedin.com/in/agemawat
YouTube youtube.com/@AdvityaGemawat
Instagram instagram.com/advitya_17
Facebook facebook.com/Advitya17
WhatsApp whatsapp.com/channel/0029VaBAHaY6buMPdgOn2R2w
