About Me
Priti Kumari
| Data Analyst & AI/ML Practitioner
I am a Computer Science student specializing in AI/ML, with experience in machine learning, deep learning, natural language processing (NLP), and large language models (LLMs).
ML/DL: Work includes building and evaluating models using scikit-learn, TensorFlow, and PyTorch, along with A/B testing and model tuning.
NLP, LLMs & Agents: Worked with Hugging Face Transformers, OpenAI models, embedding-based pipelines, and tools like LangChain and LangGraph to build LLM-driven workflows for tasks like text classification, summarization, and retrieval.
Data & Visualization: Skilled in Python, SQL, Matplotlib, Seaborn, Power BI, and Tableau for data analysis and visualization.
Prototyping: Build AI/ML-based web apps using Streamlit and Flask to demonstrate model functionality and user interaction.
Learn more about my:
Skills
Data Analysis
Python
Used for data cleaning, EDA, automation, and visualization using Pandas, NumPy, and Matplotlib.
SQL
Skilled in complex queries, joins, CTEs, and data manipulation from relational databases.
Power BI
Created interactive dashboards and reports using DAX and Power Query Editor.
Machine Learning
Scikit-learn
Used for implementing supervised and unsupervised models with preprocessing and evaluation tools.
Classification & Regression
Worked on Logistic Regression, Decision Trees, and Linear Regression in real projects.
Model Evaluation
Used ROC-AUC, Confusion Matrix, F1-Score, MAE, RMSE for model validation.
Deep Learning
TensorFlow / Keras
Built deep learning models like CNNs, RNNs, and custom architectures for image and text data.
PyTorch
Used in experimentation and academic research to design and train advanced models.
DL Models
Hands-on experience with CNNs (e.g. VGG, ResNet), UNet, and transfer learning for computer vision tasks.
Natural Language Processing (NLP)
NLP Projects
Built text classification, spam detection, sentiment analysis, and chatbot pipelines.
spaCy / NLTK
Used for text preprocessing, tokenization, lemmatization, and named entity recognition.
Transformers (Hugging Face)
Applied BERT, DistilBERT, and other models for tasks like Q&A and sequence classification.
LLMs & Agents
Working with LLMs
Used models via Hugging Face, Ollama, and OpenAI APIs for tasks like Q&A, chat, and summarization.
Embeddings & Vector Search
Worked with embeddings and FAISS for semantic search and context-aware responses.
LLM Agents
Explored agentic workflows using LangChain and custom logic to build tool-using intelligent agents.
Highlighted Projects by Domain
These aren’t just portfolio pieces — they’re stepping stones that shaped my growth from Python scripting to intelligent AI agents. I’ve grouped them by domain to showcase the range of technologies and skills applied over time.
Data Analysis Projects
SQL Business Analysis
Overview: End-to-end business insight workflow using Python and SQL. Focused on cleaning data, feature engineering, and querying to derive insights.
Skills Used:Python (Pandas) MySQL Feature Engineering
Banking Risk Analytics
Overview: Financial risk analysis using EDA and Power BI dashboards to detect loan/deposit risks. Addressed real banking case challenges.
Skills Used:MySQL Python (matplotlib, seaborn) Power BI
Software Development & Backend
Online Bus Booking System
Overview: GUI application with real-time scheduling and database handling using Python’s Tkinter and SQLite. My first large-scale application.
Skills Used:Python (Tkinter) SQLite SQL
FastAPI SQL Backend
Overview: Asynchronous API with POST/GET endpoints, Pydantic validation, and SQLAlchemy integration with MySQL. Production-ready structure.
Skills Used:FastAPI SQLAlchemy MySQL Pydantic
Machine Learning & Deep Learning
Movie Recommendation System
Overview: Built with cosine similarity on text vectors (CountVectorizer). Pulled movie posters via API and used Streamlit for UI deployment.
Skills Used:NLP cosine_similarity API Streamlit
A-B-testing
Overview : A/B testing on website conversion data to compare two UI designs; statistical hypothesis testing (Z-test), p-value analysis, and visualizations to recommend the better-performing variant.
Skills Used: Python (pandas, scipy) matplotlib statistical + practical significance.
Autism Detection in Children
Overview: Used VGG16 and InceptionV3 for image classification to detect ASD from facial features. Based on CNN pipelines and transfer learning.
Skills Used:Deep Learning VGG16 InceptionV3 Image Classification
Breast-Histopathology-Image-Classification
Overview: Built a Convolutional Neural Network (CNN) for image classification to detect Invasive Ductal Carcinoma (IDC) in breast cancer histopathology images.
Skills Used:Deep Learning CNN Architecture Image Classification
LLMs & Agentic Systems
LLM RAG PDF QA
Overview: Built a Retrieval-Augmented Generation pipeline over PDFs. Used FAISS to chunk and retrieve, with answers generated by Gemma.
Skills Used:LangChain FAISS Ollama Embeddings
Agentic AI Bot (Mistral)
Overview: Multi-agent system for intelligent generation and reasoning using Mistral 7B with prompt planning and minimal resource usage.
Skills Used:Ollama Mistral 7B LangGraph Prompt Engineering
Research
Autism Spectrum Disorder Detection
Deep learning model using VGG16 to classify Autism Spectrum Disorder (ASD) from image data.
Focused on improving early detection accuracy through model fine-tuning and dataset preprocessing.
Resume
Hobbies
Anime, gaming, building cool things with AI, writing tech blogs, collecting digital art, and exploring the future of consciousness through code