Builder at heart,
computer scientist by training.
I’m Amandeep Khurana. Endlessly curious, always pushing at the boundaries of what’s possible. I work on applied AI and AI infrastructure, advise founders, write about health & longevity, and ship things. In no particular order.
Currently at Anthropic as a Member of Technical Staff, leading Forward Deployed Engineering. Previously at AWS, where I built and shipped agent optimization capabilities in Amazon Bedrock AgentCore, infrastructure for building, running and operating AI agents. Before that, founding PM for Kiro, and CEO + Founder of Okera. Long before that, I co-wrote HBase in Action (Manning, 2012). I thrive at the intersection of technology and human potential, looking for ways to solve meaningful problems.
Beyond the day job, I’m deep into AI × Health, metabolic health, longevity, human performance, and parenting. Not just interests, but areas where I actively contribute through writing, research, and experimentation. I also advise leaders on bringing AI into their organizations, and selectively invest in startups I get excited about. Based in the San Francisco Bay Area.
§ now · what I'm up to
§ writing · notes.md
A hand-picked selection. Full archive on Substack ↗ and X ↗.
22 posts shown
- Customers Are Buying Your Future Behavior
The deals we almost lost, the systems that broke, and the customers who bet on us anyway.
- Lifestyle Inflation Is a Cage
The floor of what you need becomes the ceiling of what you can risk.
- Let Time Do the Work
The most important financial skill is the simplest one.
- Sleep Is a Superpower
Why I was picking fights about dinner at 6 PM.
- Feedback Is a Gift
The real cheat code.
- Agents of chaos, what a new study tells us about multi-agent systems
What happens when you let many agents talk to each other with no referee? A lot, and not all of it good. Reading notes and implications for production design.
- The People You Choose
What I learned about being deliberate about who I spend time with.
- Manage Your State
The earlier you learn to manage your reactions, the less they cost you.
- Supercharging engineering with Claude
My Claude Code setup, the specs, skills, and memory practices that separate demo-able agentic coding from work I'd actually ship.
- Earning or Learning
In every job, be learning or earning. Ideally both. Leave if it's neither.
- Business Leader's Brief: The Personal AI Agent Explosion
250,000 developers are building infrastructure for always-on AI agents following breakthroughs in model reasoning and tool use.
- Why health AI is a context problem
Intelligence isn't the bottleneck. The context a model has, about you, your history, your goals, is.
- The hidden cost of GLP-1s on sleep, 5 years of data
What the scale gained, the Oura ring gave back. An n=1 I wish someone had done for me before I started.
- Vibe-coding my dream health tracker
One weekend, one Apple Health export, and a dashboard I actually want to open.
- Use AI aggressively for cognition; conservatively for action.
A heuristic that's held up across teams and tools. The cheap experiments are in thinking; the expensive ones are in doing.
- Being the CEO of my health, what I track
The dashboard, the metrics, the reasoning, and the handful of things I've stopped tracking because they weren't changing behavior.
- Six months in the fog, identity and setbacks with obesity
The psychological half of metabolic disease that nobody writes a protocol for.
- South Asians and cardiometabolic disease risk
Why the default guidelines underserve 1.9B people, and what I've changed as a result.
- People don't quit decisions. They quit how decisions are made.
Notes from a founder who learned this the expensive way.
- 2018usenix ↗Modern data architecture, the deconstructed database
How the monolithic database pulled itself apart into the data stack we now take for granted.
- 2016usenix ↗Streaming systems and architectures
A survey of the streaming landscape, ca. the moment it stopped being exotic.