Practical AI, built to run in production.
I help organisations turn complex technology and AI problems into practical, production-ready solutions.
I'm a GenAI and technology architect with 21+ years across ITSM, monitoring, AIOps, enterprise architecture and IT operations. In October 2026 I was promoted to Senior Architect at Orange Business, after nine years as an Integration Specialist. My focus today is applying GenAI and LLMs to real operational problems, not AI for its own sake.
That background is practical. I've worked across SolarWinds, Cisco, Aruba, ThousandEyes, Moogsoft and Splunk, integrated them with ServiceNow, and designed AIOps solutions on AWS. It shapes how I approach GenAI: I care about adoption, reliability, cost, integration and operational value, not just what a model can do.
I like the problems that don't have an obvious answer yet. Right now that means leading a team of 20, helping drive AI adoption across 30+ client environments, and moving our observability stack toward open-source tools like Zabbix and Grafana. This blog, The Thread, is where I think out loud.
How I work
The same four steps, whatever the problem.
01
Understand the problem
Start from the operational problem, not the technology.
02
Challenge assumptions
Question the brief, the data and what stakeholders take for granted.
03
Design the right architecture
Choose what fits adoption, reliability, cost and integration.
04
Prove it
Build a working solution that shows the idea holds up.
What I work on
GenAI and LLM solution architecture
Designing GenAI and LLM solutions for real operational problems, with adoption, reliability, cost and integration in mind, not just model capability.
AI adoption and use-case discovery
Working out what is worth building before building it, and how to prove it works with a rapid prototype.
Enterprise AI and RAG architectures
Grounding LLMs in enterprise knowledge and systems so answers are useful, trustworthy and maintainable.
AI agents and IT operations automation
Agents that take on operational work such as incident triage, and sit inside the tools operations teams already use.
Observability and AIOps
Monitoring and observability across SolarWinds, Cisco, Aruba, ThousandEyes, Moogsoft and Splunk, integrated with ServiceNow and designed on AWS, now moving toward open-source stacks such as Zabbix and Grafana.
FinOps and AI-driven cost optimisation
Combining ML and LLMs to estimate and optimise cloud cost.
Architecture and solution reviews
Reviewing technology architectures, challenging assumptions, and turning ideas into production-oriented solutions.
What I've shipped
GenAI for enterprise IT operations, supporting AI-driven operations across 30+ client environments.
NOC L1 assistance
stage 01
GenAI support for network incident triage.
LLM-driven code review
stage 02
Language models reviewing code as part of delivery.
AI-powered dashboards
stage 03
Dashboard generation with automated statistics suggestions.
GenAI-powered observability
stage 04
GenAI applied to making monitoring data easier to act on.
FinOps agents
stage 05
ML and LLMs combined for cloud cost estimation.
Awards
Orange Award
Orange Business
Recognised twice at capability level for outstanding individual performance.
GEM Award
Accenture (client)
"Going Extra Mile", awarded by the client for exceptional value to deliverables.
ACE Award
Accenture
Team performance recognition at Mumbai Development Center level.
Innovation finalist
Accenture IDC
Idea selected among the final 8 entries out of 14,000 submissions.
What drives me
Technology
21 years of connecting systems that were never meant to talk to each other. It still surprises me, and occasionally infuriates me, in equal measure.
Astronomy
Looking up reminds me how small our problems really are. There's something clarifying about Saturn's rings at 2am.
Travel
Every new place recalibrates your assumptions. The best debugging sessions I've had happened while wandering somewhere unfamiliar.
Lifelong learning
The best engineers I know never stopped being students. Neither did the best managers.
Have an AI or technology problem worth solving?
If you need to understand what is worth building, how it should be architected, and how to prove it works, I'd enjoy that conversation.