Independent systems architecture practice led by Krishnakumar Avanoor. Helping organizations modernize core enterprise platforms, solve complex integration bottlenecks, build dependable data systems, and bring engineering discipline to AI agent workflows.
01 / Practice Areas
Hands-on architecture and technical consulting to solve concrete operational bottlenecks across enterprise applications, integration layers, and data systems.
Practical target roadmaps and application assessments. Helping organizations evolve established core systems (SAP, Oracle, custom applications) cleanly and incrementally, keeping operational disruption to a minimum. Also includes building Conversational BI interfaces (e.g., Databricks Genie) to make enterprise data directly accessible to decision-makers.
Connecting core systems of record with modern services using event meshes, APIs, and lightweight sidecars. Decoupling high-volume workloads—like store billing, real-time inventory checks (ATP), and near real-time pricing and promotion resolution.
Architecting modern lakehouses on open table formats (Delta Lake, Apache Iceberg) with decoupled storage and compute. Consolidating fragmented product, inventory, and operational transactions into clean, governed semantic layers across AWS, Databricks, and Snowflake.
Bringing software engineering discipline to AI-assisted development. Building spec-driven harnesses, test verification gates, and structured workflows so AI agents can be trusted on real enterprise codebases.
02 / Track Record
A track record grounded in designing, building, and operating high-throughput business platforms across retail, e-commerce, media, and finance.
AI Engineering Protocol // Oystro Technologies
Co-Founder & Architect of Oystro (oystro.com), directing product architecture for the ADLC protocol. Engineered multi-repo systems, deterministic branch protection gates, and project-graph runtimes to manage long-running agent workflows.
Enterprise Systems // Multi-Country Integration
Designed integration architecture and technical schemas bridging SAP Business One with other application platforms, alongside automated e-invoicing API blueprints across African regulatory platforms.
Conversational Analytics // Enterprise Data
Architected the migration of complex legacy Power BI models into a unified Databricks Unity Catalog semantic layer and Genie Spaces, enabling natural-language conversational analytics and automated exception monitoring.
Enterprise IT Leadership // Shoppers Stop (16 Years)
Led enterprise IT across finance (Oracle Financials), HR (PeopleSoft HRMS), e-commerce (Hybris), and master data (Riversand MDM). Implemented the enterprise Netezza warehouse with Claraview models, later migrated to an AWS data lake (S3, Redshift, EMR), and architected integration pipelines using TIBCO ESB and SAP HANA.
Retail & E-Commerce OMS // Hippostores
Architected a cloud-native, serverless OMS on AWS (SQS, Step Functions, EventBridge, Lambda) handling end-to-end order orchestration, SAP delivery synchronization, and real-time inventory ATP across high-velocity store and online channels.
Media & Transactions // HTA Fulcrum, India
Architected and delivered the enterprise Oracle platform replacing legacy FoxPro systems—handling end-to-end AOR media buying operations, release orders, ad-spot tracking, three-way bill matching, and client billing for HTA Fulcrum, India.
03 / Field Notes & Guides
Practical guides and field notes on domain modeling, systems architecture, and agentic engineering.
The 7-Part Curriculum: From Domain to Factory
Stripping away Java-era ceremony to focus on Ubiquitous Language and Bounded Contexts as ground truth.
Workshop method to map business reality with stakeholders using timeline-based domain events.
Deriving domain models and bounded contexts directly from existing enterprise documentation and swimlanes.
Using domain-ctx structures to eliminate terminology drift in LLM-assisted pipelines.
Structured execution at Levels 3 and 4: agent roles, artifact chains, and human-in-the-loop governance.
The lights-out factory model: specifications, digital twins, and holdout scenarios driving production agents.
The dependency sequence tying it all together: entering the era of deterministic Harness Engineering.
04 / Principal Profile
Enterprise Architect · Co-Founder, Oystro
I build developer tooling to bring engineering discipline to AI-assisted software lifecycles. In parallel, I take on focused architectural engagements across application modernization, integration, and modern data platforms.
My work has been shaped across core enterprise applications in retail, e-commerce, and media, focusing on the systems that keep enterprises running:
My primary engineering focus currently is on integrating agentic capabilities into established software engineering disciplines—bringing spec-driven rigor, deterministic quality gates, and structured harnesses to brownfield enterprise codebases rather than treating AI as an ad-hoc shortcut. My advisory work directly influences the tools I build, ensuring they solve real brownfield challenges, while hands-on tool building keeps my architecture practice grounded in modern execution.
As Co-Founder & Architect of Oystro (oystro.com), I am working on the tools and protocols (like the Agentic Development Lifecycle / oystro-oss) needed to manage agent-assisted software lifecycles safely. I also regularly share these learnings through practical engineering articles and guides.
Always glad to connect with fellow architects, engineering leaders, and founders to exchange notes on systems architecture, integration challenges, or practical agentic tooling.
Open to discussing systems architecture, integration bottlenecks, or sharing notes on practical agentic tooling.