AI Engineer  ·  Backend Systems  ·  LLD / HLD  ·  Agentic AI  ·  RAG Pipelines

I build production-grade AI systems that combine modern LLM capabilities with robust backend engineering. My work sits at the intersection of AI infrastructure, distributed systems, and scalable architecture.

• Designed and shipped agentic pipelines (LangGraph) for production AI applications
• Built async backend services with FastAPI + PostgreSQL + Redis
• Implemented an MCP host + intelligent tool routing layer with a Skills Protocol
• Contributed to open-source LangGraph agent hosting infrastructure (Aegra)
• Builds RAG pipelines with evaluation loops — LangFuse, LangSmith, Ragas
• Thinks in system design — LLD patterns + HLD trade-offs, not just code

Reliable AI starts with reliable systems.

🔹 MCP Host + Intelligent Tool Execution System MCP · Skills Protocol · LangGraph · FastAPI

• Integrated an MCP host as a first-class feature inside a production AI system
• Built a runtime tool discovery and routing layer that dynamically selects and executes the most relevant MCP tools for a task — no hardcoded routing
• Implemented a Skills Protocol layer: higher-level skills compose MCP tools to perform multi-step workflows
• Architecture: Task → Skill → MCP Tool(s) → Result with runtime tool resolution

🔹 Enterprise RAG Platform FastAPI · LangGraph · PostgreSQL + pgvector · Redis · Qdrant

• Hybrid retrieval (semantic + BM25) with metadata filtering
• Streaming responses over SSE
• Agent workflows with tool-use and planning
• Evaluation pipeline via LangFuse + Ragas

🔹 Agentic Workflow Engine LangGraph · Redis pub/sub · Celery · FastAPI

• Multi-agent orchestration with stateful memory
• Async background task execution with Redis as Celery broker
• Real-time status streaming to clients

🔹 Aegra — Open Source Contribution LangGraph Agent Protocol · Agent Hosting Infrastructure

• Contributed to agent hosting layer for reliability at scale
• Focused on production-grade agent serving patterns

              Client
                 │
                 ▼
         FastAPI  (REST · SSE · Auth · Rate Limiting)
                 │
     ┌───────────┼────────────┐
     ▼           ▼            ▼
PostgreSQL     Redis        Celery Worker
(ACID · jsonb  (cache ·      (background tasks)
 · pgvector)    session ·         │
                pub-sub ·         ▼
                rate limit)   LLM APIs
                                  │
                                  ▼
                             LangFuse / Eval

          User Request
               │
               ▼
         FastAPI API
               │
               ▼
       LangGraph Runtime
               │
     ┌─────────┼──────────┐
     ▼         ▼          ▼
 Retrieval   Tools      Memory
     │         │          │
     └────┬────┘──────────┘
          ▼
       LLM APIs
          │
          ▼
   LangFuse / Eval

LLD                                    HLD
──────────────────────────────────     ──────────────────────────────────────
SOLID Principles                       Distributed System Principles
GoF Design Patterns                    Horizontal Scaling + Load Balancing
Clean Architecture                     Caching (Redis — cache-aside, write-through)
Repository / Service Layer             Message Queues (Celery + Redis / SQS)
Dependency Injection                   Database Design (Normalization, Indexing, Partitioning)
Domain-Driven Design                   API Gateway + Rate Limiting
Async / ASGI Internals                 CAP Theorem & eventual consistency

AI Runtime Architecture  ·  ASGI Internals  ·  Kafka / Redpanda
Kubernetes  ·  Event-driven Architectures  ·  Distributed Systems

AI Infrastructure  ·  Agent Runtime Design  ·  MCP  ·  Vector Search  ·  Retrieval Systems