AI & Generative AI
Agentic AIGenerative AIRAGLLMsOpenAIEmbeddingsPrompt EngineeringTool CallingHuman-in-the-Loop
Ruhi Garga
AI & Technical Architect
Building intelligent, scalable systems at the intersection of AI, cloud and software architecture.
13+ years of software engineering experience
13+years engineering
Agentic AI•RAG•Generative AI•.NET•Azure•System Design•APIs•Cloud Architecture•Agentic AI•RAG•Generative AI•.NET•Azure•System Design•APIs•Cloud Architecture•
Featured Engineering Work
Two flagship case studies: a grounded enterprise RAG assistant and an agentic system with a human-approval boundary.
Ingestion
Upload documentsExtract & chunkEmbeddingsQdrant vector store
Query
Natural-language questionSemantic searchRelevant contextLLMGrounded answer
Production Engineering
JWT authCachingRate limitingAPI versioningHealth checksCorrelation IDsAudit loggingUsage tracking
AI Knowledge Assistant for Enterprises
Problem. Enterprise knowledge is spread across documents and hard to retrieve accurately. The goal: let users upload documents, ask natural-language questions and get answers grounded in that content.
Solution. An enterprise RAG application that chunks and embeds uploaded documents, stores them in a vector database, retrieves context through semantic search and uses an LLM to generate grounded answers.
- Multi-turn conversation support
- Semantic vector search
- Caching and rate limiting
- API versioning and health checks
- Structured exception handling
- Correlation IDs, audit logging and usage tracking
- Unit and integration testing
- CI/CD
Deployment. Containerised with Docker, with automated CI through GitHub Actions and Azure deployment work.
- Observe
- Gather Evidence
- Form Hypotheses
- Evaluate Readinessreadiness gate
- Propose Remediation
- Risk & Policy Evaluationdeterministic policy controls
- Human Approvalapproval boundary
- Execute
- Validate Recovery
Autonomous Production Incident Resolution Agent
Problem. Engineers must manually correlate health signals, logs, deployment history, database metrics and operational knowledge before deciding whether remediation is safe.
Solution. An agentic AI system that gathers evidence, evaluates hypotheses, proposes remediation and checks risk and policy. It pauses for human approval, executes the approved action and validates recovery.
- Evidence-based reasoning
- Readiness gates
- Deterministic policy controls
- Risk classification
- Human-in-the-loop approval
- Post-remediation validation
What I Build
Four connected areas that work together in production systems.
Professional Journey
13+ years of engineering experience — evolving from building enterprise applications to architecting intelligent, production-ready systems.
- Stage 1
Software Engineering
Building the engineering foundation13+ years of experience building enterprise software, with a strong foundation in C#, .NET, APIs and enterprise application development.
.NETC#REST APIsEnterprise ApplicationsSoftware Engineering - Stage 2
Technical Architecture
From building features to designing systemsProgressed into designing scalable and maintainable solutions, focusing on system architecture, microservices, API design, integration, security and production engineering.
System DesignMicroservicesAPI ArchitectureOAuth/SAMLScalabilitySecurity - Stage 3
Cloud Engineering
Designing for the cloud and productionExpanded architecture expertise into Microsoft Azure, cloud-native systems, containers, CI/CD, reliability and production-ready engineering practices.
AzureDockerCI/CDCloud ArchitectureObservabilityReliability - Current focusStage 4
AI & Agentic Systems
Building intelligent, governed systemsNow applying my engineering and architecture experience to Generative AI, RAG and Agentic AI, building intelligent systems that combine LLM reasoning with enterprise architecture, tools, governance and human oversight.
Generative AIRAGLLMsVector SearchAI AgentsHuman-in-the-Loop
Featured System Design
A worked URL shortener design exploring requirements, scale, data modelling, caching, partitioning, reliability and architectural trade-offs. The diagram below is an illustrative reference architecture.
Latest Writing
Ideas, architectures and lessons from building modern software and intelligent systems.
Keeping Secrets Out of Source Control: Practical Secret Detection with Gitleaks
A practical guide to detecting secrets across Git and CI/CD, adding layered controls, and responding safely when credentials are exposed.
Read Article →RAG Explained: A Beginner's Guide to Retrieval-Augmented Generation
How retrieval-augmented generation works, from documents and embeddings to a grounded answer, and when it is and isn't the right tool.
Seven Backend Patterns for Scalable Systems — and When Not to Use Them
Seven backend patterns for scalable systems, what each one solves, what it costs, and when to leave it out.
Technology Expertise
Technologies and architectural capabilities I use across AI, cloud, backend engineering and scalable system design.
Data & AI Retrieval
Vector SearchQdrantEmbeddingsSemantic SearchSQLiteRAG Pipelines
Architecture & System Design
System DesignMicroservicesDistributed SystemsREST APIsAPI ArchitectureOAuthSAMLScalabilityCachingRate Limiting
DevOps & Engineering Quality
DockerGitHub ActionsCI/CDUnit TestingIntegration TestingHealth ChecksObservabilityStructured Logging
.NET & Backend Engineering
.NETC#ASP.NET CoreWeb APIEntity Framework CoreJWT AuthenticationAPI Versioning
Azure & Cloud
Microsoft AzureAzure OpenAIAzure Container AppsCloud ArchitectureContainers
Let's build intelligent systems that work in the real world.
Open to conversations around AI, architecture, engineering and technical leadership.