I am a Software Engineer building AI-powered systems, agentic workflows, and scalable backend platforms.

Currently working on:

• Cost efficient, Scalable and Robust Agentic Workflow design patterns
• FastAPI & Distributed Backend Services
• Search & Knowledge Systems using Elasticsearch

I am deeply focused on the practical design patterns of agentic systems. Here are the core architectural challenges I am actively solving, along with the engineering blueprints I found most valuable to refer:

• Chunking & Embedding Pipelines

  • The Challenge: How do we decide on appropriate chunking and embedding strategies for different data types and sources?
  • Core Reference: LlamaIndex's Advanced Chunking and Strategy Playbook for handling heterogeneous data parsing, semantic splitting, and embedding drift.

• Retrieval Strategy Optimization

  • The Challenge: How do we design retrieval strategies that balance recall constraints with strict latency budgets?
  • Core Reference: Pinecone's System Architecture Blueprints and Arize Phoenix's RAG Evaluation Framework for optimizing hybrid search, reranking via tools like Cohere, and handling document attribution.

• 💸 Cost-Efficient Execution & Workflow Discovery

  • The Challenge: How do we prevent token bloat, optimize agentic runtimes, and automatically discover/reuse workflow patterns to minimize cost?
  • Core Reference: arXiv's Agent Workflow Optimization (AWO) Framework and GitHub's Engineering Guide to Token Efficiency for analyzing trajectory traces, compressing iterative loops into deterministic meta-tools, and utilizing semantic tool caching.

• Production Agent Evaluation

  • The Challenge: How do we evaluate agentic solutions and iteration changes for true production fit?
  • Core Reference: Anthropic’s Building Effective Agents and MLflow's LLM-as-a-Judge Evaluation Suites to transition away from static checklists toward step-wise and end-to-end evaluation metrics.

• Memory Architecture Design

  • The Challenge: How do we build appropriate memory architectures tailored to specific production use cases?
  • Core Reference: LangChain and LangGraph's State and Memory Persistence Concepts for tracking conversational buffers, semantic summary layers, and rigid token limits.

• Logging, Tracing, & Observability

  • The Challenge: How do we implement reliable logging, tracing, and monitoring across complex agent networks?
  • Core Reference: Langfuse's Distributed Trace Instrumentation Guides for tracking nested sub-agent spans, tool execution delays, and transaction graph exceptions under live user traffic.

• Agentic AI: Multi-Agent Orchestration | Self-Healing State Loops | Token-Optimized Workflows
• Information Retrieval: Advanced RAG | Hybrid Vector Search | Semantic Document Chunking | Elasticsearch
• Backend Platform: FastAPI | Distributed Background Workers | Prompt Engineering Production Primitives

• Generative AI & Agentic Workflows
• Multi-Agent Systems & Tool Calling
• Model Context Protocol (MCP)
• Advanced RAG & Vector Databases
• Semantic Search & Enterprise Search Systems
• LLM Evaluation, Prompt, and Context Engineering
• Workflow Orchestration & Knowledge Graphs

• shiksha-platform/frontend-modulefederation (#119): Implemented micro-frontend architecture solutions to scale web platform modularity.
• microsoft/STL (#2844): Engaged in low-level systems architecture design discussions regarding the standard template library.
• primer/react (#2284): Collaborated on UI framework design patterns within GitHub's official Primer design system.