Skip to content

    Subash Natarajan

    Strategy.
    Architecture.
    Delivery.

    I take enterprise AI from pilot to production. I answer the executive question, “is this safe and ready?”, then build it with their teams until they can run it.

    From writing code in 2009 to AI in production

    Enterprise AI in production · SIEL.AI

    First idea to production in 8-12 weeks; 3 of 6 pilots live

    • Previously at
    • Fujitsu
    • VMware
    • NTT DATA
    • Ericsson

    How I apply AI in a client's business

    Fix the process. Then add the agents.

    Strategy, architecture, and delivery in six steps. Each one ends in something the client keeps.

    1. 1Strategy

      Pick where AI pays first

      You get: Use cases ranked by value and readiness, a build-or-buy call, and an executive owner

    2. 2Strategy

      Map how the work really runs

      You get: A process map with the costly exceptions marked, and the baseline to beat

    3. 3ArchitectureClose-up below

      Decide step by step, system by system

      You get: A target architecture and a decision log the client's architects can challenge

    4. 4Architecture

      Set the guardrails and the platform

      You get: Agent permissions, evaluations, and an audit trail, mapped to the client's rules

    5. 5Delivery

      Prove one workflow in production

      You get: A live workflow with its test suite and cost per transaction

    6. 6Delivery

      Scale and hand over

      You get: A reusable pattern, named owners, a runbook, and before-and-after numbers

    Step 3, up close
    Every step gets one of four answersBank customer onboarding, 12 steps

    Retire

    2

    Stop doing it

    • Re-key into core banking
    • Email scans between teams

    Rule

    4

    Plain code

    • Check the form
    • Screen sanctions lists
    • Open the account
    • Send the welcome pack

    Agent

    4

    Needs judgement

    • Read ID documents
    • Chase what is missing
    • Draft the risk summary
    • Answer status questions

    Human approval

    2

    A person decides

    • Approve high-risk customers
    • Clear sanctions matches

    Rules I hold to

    • No build without a baseline
    • No go-live without evaluations
    • Cost per transaction tracked from the first slice
    • Every agent has a named human owner
    • Data stays where the client's rules say

    I think in systems, and build what the business can run.

    Every engagement starts with three questions. Who are we helping? What problem are we solving? What are we shipping this week to prove it?

    Most firms split the executive advisor from the builder. I do both: I sit with the executives who sign off, then build with the team until they can run it without me.

    From developer to Field CTO, with the dips left in2009 → now

    1. 2009

      < 1 day

      to stand up an environment, from three weeks

      Engineering and automation

      Engineer, C-DOT · Ericsson

    2. 2010s

      32

      enterprise clients migrated

      Cloud platforms

      Cloud and modernisation, NTT DATA · Logicalis

    3. 2019

      6 → 1 wk

      model release time, 21-person team

      MLOps and team leadership

      Associate Director, NTT DATA

    4. 2021

      6 mo

      two stalled bank programmes live

      Cloud-native modernisation

      Sr. Architect, VMware

    5. 2023

      28

      architects in the CoE, six markets

      AI reference architecture

      Regional Practice Lead, Fujitsu

    6. 2024 · now

      8–12 wk

      first idea to production

      Enterprise AI in production

      Field CTO, SIEL.AI