Danish Khan · Field Notes 49.28° N · --:--
Engineering leader — AI & agent platforms

Engineering leadership for intelligent systems.

Engineering leader with 15+ years building and scaling AI-native products, LLM-powered workflows, and distributed platforms in high-growth environments. I lead the Agent Platform group at Highspot, and I care about what production agents actually require — and what building them is teaching us about engineering itself.

Vancouver, BC Currently: shipping trust into agent systems
DrawingAgent platform, roughly
Drawn byD. Khan
SheetA-01
ScaleRough

Selected work

Problems, not job titles
SHT 01 / 04 Highspot 2025 – Now

Agent platform for enterprise AI — trust, governance, and execution

Lead a growing team responsible for shared agent and ecosystem capabilities: tool execution, context and memory, observability, governance, and policy enforcement — enabling three major enterprise agent families (Pydantic AI SDK, Mastra) against a common infrastructure.

→ Adopted by all three agent families within six months · 50+ tools in a policy-gated registry · 10+ architecture RFCs as shared baseline

→
SHT 02 / 04 Loblaw Digital 2022 – 2025

Search, discovery, and a catalogue platform rebuilt from the ground up

Led teams across search, catalogue, and merchandising for three commerce verticals serving 4M+ monthly users — semantic retrieval, ML re-ranking, personalization, and a greenfield multi-tenant ingestion and merchandising platform that replaced SAP Hybris.

→ Search failure rate cut ~50% · $1M+ annual savings · 30% faster merchandising turnaround

→
SHT 03 / 04 Amazon 2020 – 2022

Testing business-critical tax systems at production scale

Led 12+ engineers across tax rule-management and automated-testing platforms for business-critical financial systems. Architected TACOMA, a modular rule-management platform, and directed Hopper, a SaaS shadow-testing system that replayed production traffic against candidate changes while stubbing non-deterministic dependencies.

→ Testing cycles cut 50% · coverage improved 30% · millions of test cases in under five minutes

→
SHT 04 / 04 Morgan Stanley 2011 – 2020

Recommendations for 16,000 financial advisors

Grew from senior engineer to engineering manager building Next Best Action — an ML-powered recommendation platform that processed each advisor's book of business to surface timely, compliant client actions across Global Wealth Management.

→ 20% lift in client engagement · $200M+ attributed revenue growth

→
A working thesis
The hard part of agent systems is trust — not inference.

Models improve on their own schedule. Identity, authorization, auditability, and the judgment of when a human must stay in the loop — that's the engineering. It's also where fifteen years of building platforms turn out to matter.

On the drafting table

Published essays and works in progress

Three published; one in progress. Each starts as a question from building real systems.

Projects & experiments

Honest statuses, on purpose

Engram

Shelved · Lessons kept

A personal knowledge and productivity workspace built on structured entities, relationships, search, and agent interaction. Technically satisfying; behaviorally a failure — I built a system I admired more than I used.

Lesson: capture friction beats retrieval sophistication. Every time.

Loopsmith

Active exploration

An ongoing exploration of agentic software-development workflows: where autonomous loops genuinely help, where human judgment must stay in the loop, and what "code review" means when the author is a model.

Current question: what's the smallest durable unit of agent work worth checkpointing?

About

I'm an engineering leader and a lifelong builder. I like to understand the system, find the real constraint, simplify the problem, and help teams make sound trade-offs — close enough to the code to be useful, far enough back to see the whole board.

I grew up in India and have built my career across Mumbai, New York, Toronto, and Vancouver — through financial services, commerce, tax systems, and now enterprise AI. That route shapes how I think: different markets, different constraints, the same underlying question of what makes systems and teams hold up under complexity.

Away from work I photograph, sketch, and keep a journal that occasionally turns into poetry. I distrust confident takes — including my own — and try to write only what experience can back.

The route so far
  • 19.07° NMumbai — engineering degree, first systems
  • 40.71° NNew York — wealth management, ML platforms
  • 43.65° NToronto — commerce search at national scale
  • 49.28° NVancouver — agent platforms, present day

Elsewhere: photographs, sketches, and short reflections live in a quieter corner of this site — deliberately apart from the technical writing.