kartik.dev
P1 · service overviewservice: kartik
kartik
kartik
VP of Engineering, Platform Infrastructure & AI Core · HubSpot
STATUS
RUNNING
UPTIME
--y --m --d --:--:--
region: us-west-2 (PNW)tier: 0on-call: yes
p50 time-to-opinion
47ms-12%
p99 context switch
1.2s+3%
decisions/day
124+8%
1:1s/wk
18
error rate
0.4%-12%
meetings/wk
31+2
availability (last 90d)
placeholder series — anchor incidents arrive with /content/availability
P2 · deploy history7 deploys · oldest → newest
2010
2015
2020
2025
2026
Children's Mercy Hospital · Agfa Healthcare · Healthcare Software Engineering
2005.062011
Cerner Corporation · Principal Engineer, Brahe (Distributed Search Platform)
2011.012014
Amazon · Software Engineering Manager, Normalization (Catalog Quality)
2014.012015
Amazon · Software Engineering Manager, Catalog Fact Extraction
2015.072017
Amazon Web Services · Engineering Manager, AWS Lambda (Control Plane & Networking)
2017.122020
Block (Square) · Engineering Director, Developer Platform Services & Partner Integrations
2020.062022
HubSpot · VP of Engineering, Platform Infrastructure & AI Core
2022.05present
P3 · SLOs8 objectives · avg 88.4%

sparkline shape derived from current value + trend; per-day history lands with /content/skills/*-history.md

P4 · hard lessons6 lessons · 4 resolved · 1 ongoing · 1 wont-fix
statustitledate
P5 · career economics (indexed)base 2005 = 100
  • compensation (indexed)
  • equity (indexed)
  • learning (indexed)
  • health (indexed)
  • family (indexed)

cost optimization wins

no entries yet — populate optimizationWins in /content/cost-economics.md

P12 · weekly time allocation50h/wk · 8 categories
P6 · architecture9 services · 10 edges
P7 · dependencies4 packages
$ cat package.json
{
"dependencies": {
"feynman-lectures": "^3.0.0",
"brief-answers": "^1.0.0",
"high-output-management": "^4.0.0",
"the-pragmatic-programmer": "^20.0.0"
}
}
P8 · now

CURRENT INCIDENT (none)

CURRENT FOCUS — Q2 2026


Reliability program

  • Rebuilding the surface area on the AI-assisted incident response work — the diagnoses are accurate; the UX needs to catch up before engagement follows
  • Expanding chaos automation across the platform
  • Continuing the long arc of structural reduction in highest-tier surface area — bringing the bar down by retiring or re-tiering work, not just by adding monitors

Platform and infrastructure

  • AI Core Platform maturation — model serving, training, vector search, retrieval continuing to land under unified ownership
  • Multi-year infrastructure efficiency execution

Org

  • Promotion calibration cycle for senior engineering levels
  • Continuing the staff-level career growth program in its biannual rhythm

Personal

  • Running consistency — building back to a regular cadence
  • Reading: TODO (current book)
  • Family in the PNW; weekend hikes when the weather cooperates
last updated: 2026-05-09
P9 · log stream0 lines ·

warming up…

P11 · activity (last 3 years)124 weeks · 6 peaks
JanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecMWF
lessmore
P10 · on-call4 topics open · 0 entries
TALKS · GIVEN0 entries

no entries yet — populate talksGiven in /content/talks.md

TALKS · SCHEDULED0 entries

no entries yet — populate talksScheduled in /content/talks.md

WRITING0 entries

no entries yet — populate writing in /content/talks.md

TOPICS · ACCEPTING INVITATIONS4 entries
  • Building reliability programs that scale faster than headcount

    How to architect reliability as a system of systems with feedback loops — Prevention, Safe Deploy, Detection, Response, Validation — and why automated enforcement beats voluntary frameworks for cross-cutting concerns.

  • Backwards-compatible API evolution at hyperscale

    Lessons from introducing statefulness into a stateless API contract on AWS Lambda — async state machines, cache TTL as a consistency primitive, and how to ship a contract change for an ecosystem without breaking it.

  • AI infrastructure is infrastructure

    Why model serving, training, retrieval, and feature stores belong under the same engineering discipline that runs the rest of the platform — and what changes operationally when you consolidate.

  • Cloud vendor strategy at nine figures

    Designing the relationship, not just the contract — finding tradeable surface area, modeling commitment structures honestly, and why the best negotiation muscle is operational fluency.