---
schema: 1
type: video
title: 'Skills are the New SDKs - Elvin Aghammadzada, DataRobot'
description: >-
  You shipped a REST API. Then an SDK. Then MCP tool calling. And still, when a
  developer asks a coding agent to use your platform, it invents steps, and
  breaks in production. The problem is that your platform isn't teachable yet.
  The fix is a skill layer. Versioned, task-specific packages that encode the
  workflow knowledge your platform team has always had. Building step ordering,
  failure modes and security hygiene into something agents can load and execute.
  This became one of the hardest challenges we have seen on building enterprise
  agentic AI platform for Fortune 50. The talk walks through fundamentals of
  platform agent skills as a live case study: ML predictions, model training,
  deployment, monitoring, CI/CD, and agent observability - packaged as
  installable skills that work across Claude Code, Cursor, Codex. This pattern
  will paint a clear design pattern for building your own agent-native platform
  layer, and one question answered: what does it actually take to make your
  platform teachable? Speakers: - Elvin Aghammadzada (DataRobot): Elvin
  Aghammadzada is a data science engineer on DataRobot's Agent Workforce
  Platform, focused on bringing production-grade agentic AI to the enterprise.
  LinkedIn: https://www.linkedin.com/in/elvin-agammed/ GitHub:
  https://github.com/elvinagam
canonical: >-
  https://liquidgold.cc/channel/aidotengineer/explore/video/skills-are-the-new-sdks-elvin-aghammadzada-datarobo-adf8f93315ef
markdown: >-
  https://liquidgold.cc/channel/aidotengineer/video/skills-are-the-new-sdks-elvin-aghammadzada-datarobo-adf8f93315ef.md
llms: null
parent: 'https://liquidgold.cc/channel/aidotengineer.md'
channel: aidotengineer
updated: '2026-10-24T07:24:06.231Z'
language: null
children: []
videoId: LC3-P7v3yoI
durationSeconds: 1600
publishedAt: '2026-07-20T06:24:53.000Z'
searchChunks:
  - ordinal: 0
    text: null
    startSeconds: 3
    endSeconds: 15
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=3'
    heldReason: rights
  - ordinal: 1
    text: null
    startSeconds: 15
    endSeconds: 114
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=15'
    heldReason: rights
  - ordinal: 2
    text: null
    startSeconds: 15
    endSeconds: 114
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=15'
    heldReason: rights
  - ordinal: 3
    text: null
    startSeconds: 124
    endSeconds: 215
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=124'
    heldReason: rights
  - ordinal: 4
    text: null
    startSeconds: 124
    endSeconds: 215
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=124'
    heldReason: rights
  - ordinal: 5
    text: null
    startSeconds: 226
    endSeconds: 352
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=226'
    heldReason: rights
  - ordinal: 6
    text: null
    startSeconds: 362
    endSeconds: 476
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=362'
    heldReason: rights
  - ordinal: 7
    text: null
    startSeconds: 486
    endSeconds: 566
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=486'
    heldReason: rights
  - ordinal: 8
    text: null
    startSeconds: 577
    endSeconds: 704
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=577'
    heldReason: rights
  - ordinal: 9
    text: null
    startSeconds: 714
    endSeconds: 826
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=714'
    heldReason: rights
  - ordinal: 10
    text: null
    startSeconds: 714
    endSeconds: 826
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=714'
    heldReason: rights
  - ordinal: 11
    text: null
    startSeconds: 826
    endSeconds: 837
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=826'
    heldReason: rights
  - ordinal: 12
    text: null
    startSeconds: 956
    endSeconds: 967
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=956'
    heldReason: rights
  - ordinal: 13
    text: null
    startSeconds: 967
    endSeconds: 1075
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=967'
    heldReason: rights
  - ordinal: 14
    text: null
    startSeconds: 1086
    endSeconds: 1185
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=1086'
    heldReason: rights
  - ordinal: 15
    text: null
    startSeconds: 1197
    endSeconds: 1276
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=1197'
    heldReason: rights
  - ordinal: 16
    text: null
    startSeconds: 1276
    endSeconds: 1287
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=1276'
    heldReason: rights
  - ordinal: 17
    text: null
    startSeconds: 1374
    endSeconds: 1385
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=1374'
    heldReason: rights
  - ordinal: 18
    text: null
    startSeconds: 1385
    endSeconds: 1474
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=1385'
    heldReason: rights
  - ordinal: 19
    text: null
    startSeconds: 1484
    endSeconds: 1600
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=1484'
    heldReason: rights
  - ordinal: 20
    text: null
    startSeconds: 1484
    endSeconds: 1600
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=1484'
    heldReason: rights
  - ordinal: 21
    text: null
    startSeconds: 1484
    endSeconds: 1600
    watch: 'https://liquidgold.cc/watch/LC3-P7v3yoI?t=1484'
    heldReason: rights
---


# Skills are the New SDKs - Elvin Aghammadzada, DataRobot

You shipped a REST API. Then an SDK. Then MCP tool calling. And still, when a developer asks a coding agent to use your platform, it invents steps, and breaks in production. The problem is that your platform isn't teachable yet. The fix is a skill layer. Versioned, task-specific packages that encode the workflow knowledge your platform team has always had. Building step ordering, failure modes and security hygiene into something agents can load and execute. This became one of the hardest challenges we have seen on building enterprise agentic AI platform for Fortune 50. The talk walks through fundamentals of platform agent skills as a live case study: ML predictions, model training, deployment, monitoring, CI/CD, and agent observability - packaged as installable skills that work across Claude Code, Cursor, Codex. This pattern will paint a clear design pattern for building your own agent-native platform layer, and one question answered: what does it actually take to make your platform teachable? Speakers: - Elvin Aghammadzada (DataRobot): Elvin Aghammadzada is a data science engineer on DataRobot's Agent Workforce Platform, focused on bringing production-grade agentic AI to the enterprise. LinkedIn: https://www.linkedin.com/in/elvin-agammed/ GitHub: https://github.com/elvinagam

Canonical: https://liquidgold.cc/channel/aidotengineer/explore/video/skills-are-the-new-sdks-elvin-aghammadzada-datarobo-adf8f93315ef · Channel: [AI Engineer](https://liquidgold.cc/channel/aidotengineer.md)

> Search chunk text is held until public-source provenance is verified.

## Document chunks

> Missing: no stored document chunks are available.