{ "$comment": "Machine-readable profile for AI agents and assistants. Mirrors https://subash.io. Kept in sync with src/data/credibility.ts by src/test/agent-profile.test.ts.", "name": "Subash Natarajan", "url": "https://subash.io", "summary": "Forward deployed AI architect. Takes enterprise AI and application modernisation from pilot to production, working inside the client's teams: strategy with the business, architecture with the delivery teams, code in the repository.", "currentRole": { "title": "Field CTO", "organisation": "SIEL.AI", "since": "2024" }, "basedIn": [ "Singapore", "London" ], "regions": [ "APAC", "EMEA" ], "experienceSince": 2009, "contact": { "email": "suboss87@gmail.com", "linkedin": "https://www.linkedin.com/in/subashn/", "github": "https://github.com/suboss87" }, "bringInFor": [ "Shaping an AI or modernisation programme before it is signed, then delivering it", "Taking an AI pilot to production in a regulated environment", "Modernising legacy applications with agents and moving them to cloud", "Re-engineering a process before automating it", "Advising leadership on AI, cloud, and modernisation" ], "notAFitFor": [ "Consumer and mobile products", "Blockchain and Web3", "Fundamental ML research" ], "journey": [ { "since": "2024", "layer": "Enterprise AI in production", "role": "Field CTO", "company": "SIEL.AI", "proof": "First idea to production in 8-12 weeks; 3 of 6 pilots live" }, { "since": "2023", "layer": "AI reference architecture", "role": "Regional Practice Lead", "company": "Fujitsu", "proof": "Cloud CoE grown to 28 architects across six ASEAN markets" }, { "since": "2021", "layer": "Cloud-native modernisation", "role": "Sr. Architect", "company": "VMware", "proof": "Two stalled banking programmes into production within six months" }, { "since": "2019", "layer": "MLOps and team leadership", "role": "Associate Director", "company": "NTT DATA", "proof": "Led a 21-person global team; model releases cut from six weeks to one" }, { "since": "2010s", "layer": "Cloud platforms", "role": "Cloud and modernisation", "company": "NTT DATA · Logicalis", "proof": "32 enterprise clients moved onto an NTT DATA cloud platform" }, { "since": "2009", "layer": "Engineering and automation", "role": "Engineer", "company": "C-DOT · Ericsson", "proof": "Automated provisioning: new environments in under a day, down from three weeks" } ], "method": { "principle": "Re-engineer the process, then add the agents. Every step of a process goes into one of four buckets: retire, rule, agent, or human approval. An agent gets only the steps that need judgement.", "steps": [ { "step": "Agree the outcome and the baseline", "output": "A one-page value case with the numbers to beat" }, { "step": "Map how the work really runs", "output": "A process map with the costly exceptions marked" }, { "step": "Decide step by step, system by system", "output": "A target architecture and a decision log the client's architects can challenge" }, { "step": "Prove one workflow in production", "output": "A live workflow with its test suite and cost per transaction" }, { "step": "Set the foundation, then scale", "output": "A reusable pattern and a roadmap ranked by value" }, { "step": "Hand over and keep measuring", "output": "Named owners, a runbook, and before-and-after numbers" } ], "rules": [ "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" ] }, "engagements": [ { "sector": "Manufacturing", "title": "Accounts payable, re-engineered", "results": [ "Cycle time per invoice 18 to 3 minutes", "Cost per invoice $7.58 to $2.34", "Field accuracy 65% to 98%" ], "volume": "6,800 invoices a month" }, { "sector": "E-commerce pilot", "title": "A support agent that can afford to scale", "results": [ "97% fewer model calls", "Agents per conversation 6 to 1" ] }, { "sector": "Regulated environment", "title": "34 RPA bots rebuilt behind the firewall on a local LLM layer", "results": [ "Monthly maintenance 55 to 14 hours", "Cases needing a person 40% to 18%" ] }, { "sector": "Knowledge assistant", "title": "Policy search with permission-aware retrieval", "results": [ "Time to find an answer 25 to 3 minutes", "Design reused in 2 later builds" ] }, { "sector": "Banking (in progress)", "title": "On-premises agentic AI for a Gulf bank under SAMA rules, with customer data kept inside the bank", "results": [] } ], "clientsNamed": false, "recognition": [ "InfoQ Best Article 2025 (co-author, Agentic AI Architecture Framework for Enterprises)", "Co-author with Jeveen Jacob, Multi-Cloud Handbook for Developers (Packt, 2024)", "Contributing author, The New Stack", "Certified: TOGAF, Google Cloud, Kubernetes" ], "writing": [ { "title": "Agentic AI Architecture Framework for Enterprises", "outlet": "InfoQ", "url": "https://www.infoq.com/articles/agentic-ai-architecture-framework/", "role": "Co-author" }, { "title": "Multi-Cloud Handbook for Developers", "outlet": "Packt", "url": "https://www.packtpub.com/en-us/product/multi-cloud-handbook-for-developers-9781804618707", "role": "Co-author with Jeveen Jacob" }, { "title": "Practical notes on forward deployed engineering", "outlet": "fdeops.io", "url": "https://fdeops.io/" }, { "title": "The Rise of Intelligent Cloud", "outlet": "The New Stack", "url": "https://thenewstack.io/the-rise-of-intelligent-cloud/" }, { "title": "Architectural Considerations for Creating Cloud Native Applications", "outlet": "The New Stack", "url": "https://thenewstack.io/architectural-considerations-for-creating-cloud-native-applications/" } ], "previouslyAt": [ "Fujitsu", "VMware", "NTT DATA", "Ericsson", "Logicalis", "C-DOT" ], "recommendations": "https://www.linkedin.com/in/subashn/details/recommendations/", "lastUpdated": "2026-10-22", "alsoKnownAs": [ "Subash N" ], "aiReadinessReview": { "url": "https://subash.io/assessment", "price": "Complimentary", "capacity": "Three companies a month", "for": "Teams outside software that need to adopt AI but lack the budget for a consultancy or the right expert; also teams with stalled pilots", "steps": [ "Application (3 minutes); decision within 3 working days", "Preparation survey (15 minutes)", "45-minute working session", "One-page written read within 5 working days" ], "readCovers": [ "Maturity stage on a four-stage scale from MIT research", "The one workflow to start with and its cost today", "Build, buy, or defer for each part", "The next 90 days, with owners" ] } }