hey, I'm Akash 👋
Welcome! I'm an adventurer at heart ❤️ and a technologist by profession. I currently work as the Lead Technical Consultant — Partner at 
Bloomreach Inc. — helping partners build better, move faster, and get the most out of AI-driven commerce.
Growing up in multiple cities across 🇮🇳 India, thanks to my family's army background, I've developed a knack for adaptability and a love for exploring new places. Now calling 🌉 San Francisco home.
what makes me tick?
I picked up snowboarding not too long ago and I'm already obsessed. There's something about the mountain that resets everything — no emails, no Slack, just speed and snow. Still learning, still falling, fully hooked. 🏂
I cook a lot — mostly Indian food, sometimes experiments that go sideways. Badminton is my go-to for switching off the brain. Fast game, instant results, very different from the slow burn of enterprise software. 🍳🏸
Punjabi music is always on — check out my 
Spotify playlist if you want to know exactly what I'm listening to while building workflows at midnight. 🎵
things I've built
A mix of professional work, side projects, and things built out of curiosity.
see all projects →thinking about
Things I'm noodling on — not conclusions, just where my head is.
Everyone's racing to build the most capable agent. The real gap I keep seeing is evaluation — nobody really knows how to measure whether an agent did the right thing, only whether it finished the task. Capability without judgment is just automation with extra steps.
Putting LLMs into implementation workflows sounds good until the first hallucination lands in a production config. The hard part isn't getting AI to help — it's knowing exactly where to trust it and where a human still needs to be in the loop. That line is different for every client.
Personalization used to mean showing the right product. Now it's about orchestrating the right experience — across search, content, and conversation, in real time. The vendors who get this are building differently. The ones who don't are just putting an AI badge on a recommendation engine.
The interesting problem with agentic orchestration isn't the AI part — it's the coordination layer. Who owns the handoff between agents? What happens when one stalls mid-flow? The patterns that work in single-LLM setups start to break down fast when you chain four of them together.