I build:

• Retrieval-Augmented Generation (RAG) systems
• AI-powered information retrieval pipelines
• Machine learning prediction models
• AI backend services

AI-powered personalized news retrieval system using LLM-based agent, query expansion, web search, credibility filtering, and summarization.

User Email + Query
↓
Fetch User Interests (SQL)
↓
AI Agent (Tool Calling)
↓
LLM Query Expansion  
↓  
DuckDuckGo Search  
↓  
Credibility Filtering  
↓  
News Summarization

• Python
• Groq LLM API
• DuckDuckGo Search
• SQL

Retrieval-Augmented Generation system for answering medical queries using domain documents.

Medical Documents
↓
Text Chunking
↓
Embeddings (Sentence Transformers)
↓
Vector Database (Chroma)
↓
Semantic Retrieval
↓
LLM Answer Generation

• PDF ingestion and chunking
• Embeddings and semantic search
• Vector database retrieval
• LLM-based answer generation

• Python
• LangChain
• ChromaDB
• Sentence Transformers
• LLaMA / Mistral

Machine learning model predicting turbine failures using 40+ sensor variables.

Sensor Data
↓
Data Cleaning
↓
Feature Scaling
↓
Neural Network Model
↓
Failure Prediction

• Improved minority-class recall 32% → 92%
• Achieved 95% overall accuracy
• Addressed severe class imbalance

• Python
• Scikit-learn
• Neural Networks
• Pandas / NumPy

Computer vision model detecting helmet usage using VGG16 transfer learning.

Image Dataset
↓
Data Augmentation
↓
Transfer Learning (VGG16)
↓
CNN Training
↓
Helmet Classification

• 96.8% validation accuracy
• Implemented data augmentation and transfer learning

• Python
• TensorFlow
• CNN
• VGG16

Python, SQL

Scikit-learn
Feature Engineering
Model Evaluation
Classification / Regression

TensorFlow
PyTorch
CNNs
Transfer Learning

RAG Pipelines
LangChain
Embeddings
Prompt Engineering
Query Expansion

FastAPI
Git
Streamlit

I also contribute to open source using another account:
https://github.com/sakeena-7878