AI/ML Engineer Open Source Contributor
Building production-grade multi-agent systems and AI infrastructure. Developing scalable ML pipelines and autonomous reasoning systems for real-world problems.

I architect and ship multi-agent systems, distributed AI infrastructure, and production ML pipelines. I work with standardized protocols like MCP to enable agent interoperability, optimize inference, [...]

• Multi-Agent Systems & MCP: Build Agent-to-Agent (A2A) communication patterns, implement Model Context Protocol (MCP) for standardized tool interoperability, design agentic loop architectures
• AI Infrastructure: Develop distributed inference systems, retrieval pipelines, and real-time data systems with latency optimization and production robustness.
• Production ML: Ship end-to-end ML systems with evaluation frameworks, automated deployment, and monitoring strategies.
• Developer Tools: Build CLI tools and IDE extensions that bring AI capabilities directly to developers.
• Open Source: Contribute to production AI systems, build reusable frameworks, and maintain code for real-world usage.

Python • Multi-Agent CLI • VS Code Extension • Gemini • Claude • Copilot

• What I Built: One-click AI agent launcher directly from VS Code. Integrated Gemini, Claude, and Copilot into a seamless CLI experience. The fastest way to access multiple AI models from your terminal.

• Technical Implementation:

  • Multi-agent orchestration for different LLM providers
  • VS Code extension integration with CLI commands
  • Environment-based configuration for API keys
  • Async task execution for responsive UX
  • Real-time streaming responses

• Results: Production-ready developer tool with comprehensive docs and active development. Deployed to users for daily AI agent access.

Python • NVIDIA NIM • ChromaDB • SerpAPI • Gradio • FastAPI • CLI

• What I Built: Professional, modular hybrid Retrieval-Augmented Generation (RAG) system using NVIDIA Llama-3.3 LLM with real-time web search (SerpAPI) and persistent local vector database (ChromaDB).

• Technical Implementation:

  • Intelligent Query Routing: LLM-based router that automatically classifies queries between LOCAL (documents), WEB (search), or NONE (conversation)
  • Hybrid Retrieval Pipeline: Integrates SerpAPI for live web search with ChromaDB vector database using all-MiniLM-L6-v2 embeddings
  • Multi-Format Ingestion: Supports .txt and .pdf files with automatic chunking, embedding, and persistent storage
  • Stateful Memory Management: Maintains rolling conversation history buffer for multi-turn dialogue context
  • Error Handling & Fallback: Graceful degradation with automatic fallback to general knowledge on missing keys or empty databases
  • Flexible Deployment: Gradio web UI (http://127.0.0.1:7860), lightweight CLI, and FastAPI REST server (http://0.0.0.0:8000)
  • Comprehensive Testing: 13+ unit tests (mocked) + integration tests with live API validation
  • Production Architecture: Modular src/ structure, environment-based configuration, no hardcoded secrets

• Results: Production-ready RAG system with comprehensive documentation, multi-mode deployment (web/CLI/API), 13+ test suite, and demonstrated handling of edge cases like fallback routing and memory management.

Python 3.12+ • Pydantic • Async/Await • Google ADK • MCP • Agent-to-Agent Protocol

• What I Built: Enterprise-scale multi-agent system for GitHub repository analysis. Implemented primary orchestrator agent that dynamically spawns specialized Debug Agents via A2A protocol. Integrated MCP specification.

• Technical Implementation:

  • Implemented MCP Specification: Built MCP server for standardized tool/resource sharing between orchestrator and specialist agents, enabling interoperable multi-agent workflows
  • Agent Delegation System: Designed A2A protocol for agent spawning, task assignment, and result aggregation with proper error handling and retry logic
  • Multi-LLM Support: Built abstraction layer supporting OpenAI, Google Gemini, local Ollama/vLLM with unified interface
  • Production Architecture: Strict typing with Pydantic, async/await concurrency, CLI interface, environment configuration, structured logging
  • Tool Integration: Integrated GitHub API with MCP-compliant tool protocol for code analysis, bug detection, security scanning

• Results: Deployable agent framework with extensible architecture. Demonstrated sophisticated coordination patterns and standardized protocol implementation.

Python • Streamlit • Pandas • Prophet • Real-time Data Analysis

• What I Built: Interactive Streamlit app for real-time stock analysis and AI-driven price forecasting. Features live market data, sector-wise stock charts, technical indicators, news sentiment analysis.

• Technical Implementation:

  • Real-time market data integration
  • Sector-wise filtering and dynamic charting
  • Technical indicator calculations
  • News sentiment analysis for market context
  • Prophet time-series forecasting for price predictions

• Results: Full-stack analytics tool with customizable date ranges, sector selection, and chart types for informed investing decisions.

• Agent Protocols & Frameworks: Model Context Protocol (MCP), LangChain, Pydantic, Google ADK, Ollama, vLLM
• Tools & Libraries: NumPy, Apache Spark, functools, async/await, Prophet, SerpAPI
• Vector Databases: ChromaDB, Pinecone, FAISS
• Evaluation & Monitoring: RAGAS, MLflow, structured logging, error tracking

• 🚀 Credly Certifications — AI credentials
• ☁️ Google Skills Profile — Google training
• 👨‍💻 OpenAI Developer Community — Active member
• 🤖 Google Developer Group — Community contributor

Seeking internship or junior engineer roles in AI/ML systems engineering. Open to roles focused on multi-agent systems, inference infrastructure, developer tools, or production ML platforms.

Connect on LinkedIn | View my repositories