Pinned Medium
graph raggnnretrieval
The Missing Piece in Graph RAG: Graph Attention Networks
A pinned GraphRAG essay on using graph attention to make retrieval context more query-sensitive for multi-hop reasoning.
Technical blog
Technical Blog
Medium essays and local notes on AI agents, inference, EvalOps, data integration, AI UX, and startup AI systems.
agent frameworkagentsai infrastructureai uxarchitectureazureazure ai searchcontext engineeringdatadecision-makingdeploymentdurable executionevaluationgenerative uignngpugraph ragguardrailsinferenceinterfaceslatencyllmsmetadatamicrosoft buildobservabilitypersonalizationprivacyproductproduction airagreasoningretrievalsearchsecuritystartupsworkflows
Pinned
Highlighted From Medium
Pinned TDS Archive
agentsdecision-makingsearch
Tackle Complex LLM Decision-Making with Language Agent Tree Search (LATS) and GPT-4o
A pinned agent-decision essay on Language Agent Tree Search, deliberation, and inference-time search for harder LLM tasks.
Pinned Medium
ragazure ai searchmetadata
Boost RAG Performance: Enhance Vector Search with Metadata Filters in Azure AI Search
A pinned Azure AI Search essay on metadata filters as a practical retrieval control for production RAG systems.
Latest
Published Posts
Build Signal 2026
microsoft buildagentsai infrastructure
Build 2026 AI Builder Atlas
Intent: map Microsoft Build 2026 AI announcements into builder-ready samples across model, context, tools, harness, and infrastructure.
Microsoft Azure
agentsazuredurable execution
Building Agent Harnesses with Microsoft Agent-Framework Durable Extensions
How durable runtime contracts bound long-running agent behavior with state, approvals, checkpointing, and recovery.
Microsoft Azure
ragretrievalproduction ai
10 RAG Shifts Redefining Production AI in 2026
A systems view of RAG moving toward composable retrieval, late interaction, graph reasoning, freshness, and orchestration.
Microsoft Azure
agentscontext engineeringazure
Context Engineering with Microsoft Agent Framework's Context Provider API
Context engineering as a bounded prompt-assembly control surface for short-term memory, persistent memory, retrieval, and policy.
Medium
inferencelatencygpu
Stop Blaming the Model: Topological Hardening for Predictable Inference Latency
Inference latency as a topology problem: placement, routing, queues, batching, and system-level control around model serving.
Microsoft Azure
ragazure ai searchretrieval
Fixing Sparse Retrieval with RAPTOR on Azure AI Search
A practical RAPTOR pattern for sparse retrieval when relevant context is distributed across many documents.
Medium
personalizationreasoningllms
Reasoning over Features, Not Identities for Personalisation
Notes on using features and segmentation as safer inputs for reasoning-based personalization systems.
Microsoft Azure
agentsagent frameworkproduction ai
Building Production AI with Microsoft's Agent Framework: Credit Underwriting Case Study
Typed workflow graphs, multi-agent fan-out, HITL, telemetry, and UI feedback loops in a credit-underwriting demo.
agentsworkflowsobservability
Production Agent Workflows: Orchestration and Observability
Typed workflow graphs, fan-out/fan-in execution, checkpointing, human approval gates, and telemetry for production-grade agent systems.
Medium
ai uxgenerative uiinterfaces
Generative UI
Adaptive interface notes on turning user intent and engagement context into typed UI schemas that remain governable.
securityprivacyguardrails
Local PII Pre-Filter with Microsoft Presidio and Qwen 2.5
A local pre-prompt guardrail using Presidio and a small CPU model to reduce PII exposure before LLM calls.
dataagentsarchitecture
AI Paradigm Shifts and Their Data Requirements
Notes on why agentic AI, assisted coding, and AI governance require unified and organized data foundations.
agentsstartupsproduct
Forward Deployed Engineering for Building AI Agents
Why useful AI agent products often require domain workflow discovery and deployed engineering loops.
agentsazuredeployment
Building Agents on Azure
Implementation notes on Azure agent building blocks, deployment tradeoffs, and production-oriented agent patterns.
Cloud Atlas
ragretrievalevaluation
How to improve RAG peformance — Advanced RAG Patterns — Part2
A practical Advanced RAG follow-up covering data preparation, embeddings, retrieval strategy, synthesis, and prompt conditioning.
Cloud Atlas
ragretrievalproduction ai
Why do RAG pipelines fail? Advanced RAG Patterns — Part1
A failure-mode map for RAG systems across retrieval, augmentation, and generation, with concrete examples of where pipelines lose relevance.