About
Hi, I am Subhash Dasyam.
I work at the intersection of Security Architecture, DevSecOps and Generative AI, helping enterprises design systems that are both secure and practical to run in the real world.
I am currently based in Dubai, UAE, working as a Senior Manager / Lead Security and GenAI Architect in banking. My everyday work lives where containers, cloud, GPUs, and regulations all meet. I spend a lot of time translating complex security and AI concepts into something that CIOs, CISOs, architects and developers can all agree on.
What I Do
Enterprise Security Architecture
-
Design and review architectures for large scale platforms, especially on Red Hat OpenShift, Kubernetes and hybrid cloud.
-
Define security patterns, guardrails and reference architectures that teams can reuse instead of reinventing security for every project.
-
Work closely with platform, application and infrastructure teams to make sure security is built in, not bolted on.
DevSecOps and Platform Security
-
Bring DevSecOps practices into real pipelines with tools like Jenkins, container registries, SAST, DAST, software composition analysis, and container scanning.
-
Help teams move from traditional workloads to containers and virtualization while keeping controls, firewalls and IP schemes consistent with regulatory expectations.
Generative AI and Agentic Systems
-
Design GenAI solutions that respect banking regulations, privacy and data residency.
-
Work on RAG architectures, prompt engineering and AI governance so that AI outputs are traceable, explainable and auditable.
-
Build and experiment with agentic AI systems that can orchestrate tools, search, code and infrastructure in a controlled way.
Things I Have Built and Explored
I enjoy moving between strategy and hands on work. Some of the areas I have been exploring recently:
-
Automated Threat Modeling for architecture teams using LLMs, diagrams and project context.
-
AML and Sanctions Intelligence using RAG, internet research, and careful scoring frameworks for predicate offenses.
-
Agentic Red Teaming suites that simulate realistic attack paths on LLM based systems.
-
Voice based assistants using local models on devices like Jetson Orin Nano, with Whisper, TTS engines and wake word listeners.
-
GPU focused experimentation with local models, vLLM, Triton, Ollama and other inference stacks.
How I Think
I believe good security and good AI systems share the same principles:
-
Clarity over hype
If we cannot explain how a system works, we probably should not deploy it in a bank. -
Patterns over one off solutions
A strong pattern saves dozens of teams from repeating the same mistakes. -
Evidence over assumptions
Logs, metrics, test cases, TEVV style validation and risk registers matter as much as architecture diagrams. -
Continuous learning
AI and security move fast. I treat everything as a versioned artifact that can be refined: prompts, policies, patterns and playbooks.
Background and Journey
I have worked across:
-
Enterprise banking in Dubai, focusing on platform and security architecture.
-
High growth tech environments in Canada as a DevSecOps / Principal DevSecOps Engineer, working with cloud platforms like GCP and AWS, Kubernetes, and modern CI/CD.
This mix of highly regulated banking and startup style execution helps me design solutions that are innovative but realistic for production.
Outside of Work
Outside work, I enjoy:
-
Exploring AI use cases that make life easier at home and at work: personal assistants, meeting summarizers, document workflows and more.
-
Planning family friendly travel that balances experiences, comfort and logistics.
-
Writing and sharing thoughts on containers, security, AI and architecture on my blog and social channels.
Connect With Me
I am always open to thoughtful discussions on security, Gen AI, DevSecOps and architecture.
🔗 LinkedIn: linkedin.com/in/subhashdasyam
💻 GitHub: github.com/subhashdasyam
🌍 Website: www.subhashdasyam.com