I build production-grade AI systems — from autonomous LLM agents to scalable MCP server architectures. Currently working as a Machine Learning Engineer at Pivotly (Remote, MN, USA), designing intelligent agents that reason over structured databases, vector embeddings, and cloud storage at scale.
- 🔭 Currently building: Custom MCP Servers + LangGraph-based Human-in-the-Loop AI Agents
- 🌍 Working with: Azure AI Studio, Azure Document Intelligence, Azure AI Vision
- 📄 Published research in IEEE and Computers and Electronics in Agriculture (Elsevier)
- 🎓 B.Sc in CSE, Major in Data Science — CGPA: 3.87/4.00
- 🤝 Open to collaboration on AI Agents, MCP integrations, and LLM-powered systems
Languages
AI / LLM
ML / DL Frameworks
Backend & DevOps
Built local and remote Model Context Protocol (MCP) servers using FastMCP — includes expense tracking with SQLite, tool orchestration, and production-ready agent integration patterns. Covers both stdio and SSE transport layers for flexible deployment.
FastMCP SQLite Python LLM Agents
Locally hosted multi-model LLM platform running DeepSeek R1, Llama 3.2, Mistral with Streamlit interface — deployed at mlhub.space using Docker + Nginx + RTX 3060 GPU. Reduced latency by 25%, accelerated deployment by 30%.
Streamlit Ollama Docker Nginx Cloudflare
AI-powered Excel analysis app with interactive querying over structured data — containerized and deployed within the mlhub.space network. Supports multi-sheet reasoning and natural language data exploration.
Streamlit LangChain Docker Python
Autonomous financial analysis agent with tool-use capabilities — performs real-time stock retrieval, web search, and multi-step reasoning to generate structured investment insights.
LangChain Agentic AI FastAPI Python
| # | Title | Venue | Year |
|---|---|---|---|
| J1 | Accurate water level monitoring in Alternate Wetting and Drying rice cultivation using attention-based ConvNeXt architecture | Computers & Electronics in Agriculture (Elsevier) | 2025 |
| C2 | Attention-based feature fusion for Monkeypox skin lesion detection | IEEE ICCIT | 2023 |
| C1 | Implementing Federated Learning based on RainForest Model | IEEE I2CT | 2023 |
| T1 | Enhanced speech emotion recognition with efficient channel attention guided deep CNN-BiLSTM framework | arXiv | 2024 |
"Building AI systems that don't just work in notebooks — but work in production."