I’m an AI Engineer focused on building practical AI applications using Python, Generative AI, RAG, and Agentic AI.
I build end-to-end systems spanning LLM applications, retrieval pipelines, AI agents, machine learning, and deep learning - from data processing and model development to deployment and user-facing applications.
I’m currently focused on building production-oriented AI systems and contributing to open-source projects.
I particularly enjoy solving concrete problems with neural networks and building intelligent systems that create measurable impact.
- Python
- SQL
- Large Language Models (LLMs)
- Retrieval-Augmented Generation (RAG)
- AI Agents
- Multi-Agent Systems
- Prompt Engineering
- Embeddings
- Vector Databases
- Local LLMs
- Classification
- Regression
- Clustering
- Feature Engineering
- Model Evaluation
- Hyperparameter Optimization
- TensorFlow
- Keras
- CNNs
- Transfer Learning
- NLP
- Computer Vision
- LangChain
- FAISS
- Ollama
- Groq API
- Streamlit
- Git & GitHub
- Jupyter Notebook
- VS Code
- MySQL
A domain-specific conversational AI assistant built with Retrieval-Augmented Generation (RAG). The system retrieves relevant information from a FAISS vector database using Gemini embeddings and uses a Groq-hosted LLM to generate grounded responses based on the retrieved context.
The application is built with LangChain and Streamlit, demonstrating an end-to-end RAG pipeline from document ingestion and vector indexing to semantic retrieval and LLM-powered response generation.
Technologies: Python, LangChain, FAISS, Gemini Embeddings, Groq, Streamlit
Developed an AI-powered financial research agent that combines deterministic query routing, financial data retrieval, web research, and LLM-based synthesis to generate structured company research reports.
Technologies: Python, Phidata, Groq, GPT-OSS-20B, Yahoo Finance, DuckDuckGo (DDGS), Streamlit
Designed an NLP pipeline to discover and group related news events using sentence embeddings and unsupervised learning techniques. Generated event timelines and automated labeling for improved information discovery.
Technologies: Python, NLP, Sentence Embeddings, Scikit-learn, Clustering
Built a deep learning system for retinal image classification using EfficientNetB0 and TensorFlow.Applied transfer learning and class weighting to improve minority-class detection and deployed the model through a Streamlit application.
Technologies: Python, TensorFlow, EfficientNetB0, Streamlit
- Resolved an analytical error in the custom surface normal computation and corrected the corresponding tutorial implementation.
- Updated the mathematical formulation and visualization to keep the documentation consistent with the corrected implementation.
- PR #752 — Merged
- Generative AI & LLM Applications
- Agentic AI & Multi-Agent Systems
- Retrieval-Augmented Generation
- AI Application Development
- Machine Learning & Deep Learning
- NLP & Computer Vision
- AI Deployment & MLOps
Oct 2025 – Present
Working on practical machine learning, deep learning, Generative AI, RAG, and AI agent applications using Python and modern AI frameworks.
LinkedIn: https://www.linkedin.com/in/praband-kumar/
Email: praband10@gmail.com