I work at Microsoft , as part of the Applied AI Engineering team within Core AI . I lead a team of researchers, scientists, and engineers responsible for building cutting-edge AI products and services that power both Microsoft (other product and engineering teams) and our customers globally.
The areas I cover span the breadth of offerings in Azure Foundry from Cognitive Services (Vision, Speech, Language), Frontier models Azure OpenAI (GPT 5.1, GPT 5.0, GPT-4, GPT-4o, GPT 3.x (remember those?), DALL-E, Copilot, etc), Azure Machine Learning (AML), and Agentic systems and RAI. I’m responsible for leading custom engineering efforts with strategic customers, bringing world-first AI technologies and research innovations to production at scale.
My work centers on making advanced AI accessible and practical for enterprise applications:
Model Customization & Adaptation - Leading efforts in fine-tuning, model adaptation, and customization techniques spanning from mid-training, post-training, and fine-tuning. These include supervised fine-tuning (SFT), direct preference optimization (DPO), low-rank adaptation (LoRA), and reinforcement learning from human feedback (RLHF). I help customers adapt foundation models to their specific domains, data, and use cases.
Large Language Models (LLMs) & Small Language Models (SLMs) - From GPT-4o to the Phi family of models, I work across the full spectrum of language models. I’m particularly passionate about small language models that deliver capable performance on edge devices and resource-constrained environments while maintaining privacy and reducing costs.
Multi-Agentic Systems - Building and deploying AI systems where multiple agents collaborate to solve complex problems, leveraging reasoning capabilities and orchestration patterns for enterprise workflows.
AI Foundry & Platform Engineering - Managing AI Foundry’s model customization capabilities, ensuring enterprises have the tools and infrastructure to build, customize, evaluate, and deploy AI models at scale with proper governance and safety controls.
Custom Engineering & Innovation - Working directly with key customers on cutting-edge engineering challenges, collaborating with Microsoft Research (MSR) and distributed global teams to bring research breakthroughs to production systems serving millions of users.
Research & Contributions - I contributed to the Phi-3 Technical Report (arXiv:2404.14219), which introduced the Phi-3 family of small language models demonstrating that smaller models can rival larger ones while remaining deployable on edge devices - critical for privacy-preserving, on-device AI applications.
I’ve authored multiple books and other publications; see the full list on the Publications page.
About This Blog
This is my personal blog where I share thoughts on my professional experiences in AI, machine learning, and technology. While most entries are technical, sometimes they venture into varied topics—occasionally wacky, rarely political. These are my personal perspectives and don’t necessarily reflect Microsoft’s views.
About the blog’s name: Curious about the “gibberish” and the funny title “Desigeek”? Some have even joked it’s a virus that infected their system (ahem 😏). It’s quite simple, and if you haven’t figured it out yet, check out what’s in a name?
Depending on the flavor of the day, including what drama I am either brewing or working through I might be on some social media where I share insights on AI trends, research, and industry developments:
X (Twitter) : @bahree
Mastodon : @[email protected]
LinkedIn: Amit Bahree
Feel free to reach out—I love connecting with fellow technologists, researchers, and builders interested in AI’s potential and challenges.
My Mastodon Introduction Tooted on Hi, my name is Amit. I work at #Microsoft in the #ai platform team - we build all the AI products that are used by our customers and internally by other product and engineering teams. I am responsible for custom engineering on the platform and span across the platform - including #Azure Cognitive Services, Azure OpenAI (#GPT3, #DALLE.2, #Codex, #Copilot), AML, etc. I am a simple geek who loves to tinker and wallow in code. At home the dog rules after the wife. 🤓 🐶
Recent Thoughts
View on Twitter
A bit more about me: I’m a geek who loves to tinker and wallow in code. At home, the dog rules (after the wife 🤓🐶). When I’m not working, you’ll find me reading AI research papers, writing technical blog posts, or building custom LLM projects exploring fine-tuning techniques, model adaptation, and AI infrastructure challenges.