Kriti C Parikh

Remembri is a lightweight, logic driven memory system that enables LLMs to remember user specific facts—like name and location, without embeddings or chat history. Built with FastAPI and SQLite, it reduced repeated input by 73%, achieved 95% tool-calling accuracy, and boosted context relevance by injecting symbolic memory directly into prompts.

• View project

Built a real-time AI wellness assistant using Hugging Face and Gradio to generate personalized meal/workout plans and detect user emotions through NLP. Integrated Tavily API for smart-search research bot, enabling real-time wellness recommendations with visual and contextual outputs.

• View project

Built a multi-agent AI teaching assistant using LangGraph, Vertex AI, and Gemini LLMs to automate lesson planning, quizzes, and personalized learning. Implemented an event-driven architecture with Pub/Sub and Eventarc to boost agent collaboration and real-time content generation. Integrated RESTful APIs via Cloud Run to deliver scalable, multimodal educational content, developed through an interactive course and direct guidance from Googlers.

• View project

This project leverages Tableau to analyze and visualize key aspects of electric vehicle (EV) data, with a focus on Battery Electric Vehicles (BEVs) and Plug-in Hybrid Electric Vehicles (PHEVs). The interactive dashboard uncovers trends in the EV market, highlights geographical distribution, and evaluates manufacturer performance.

• View project

Developed an interactive and visually intuitive Power BI dashboard to monitor website performance metrics. The dashboard was tailored for stakeholders with non-technical backgrounds to make data-driven decisions effectively.

• View project

This project applies time series analysis to two datasets—Walmart sales (seasonal) and Formula 1 results (non-seasonal)—using the Box-Jenkins methodology to uncover patterns, forecast trends, and enhance decision-making.

• View project

This project is a sophisticated tool that allows users to interact with multiple PDF documents through an AI-powered chat interface. Leveraging advanced technologies like Gemini AI and LangChain, the system extracts, processes, and stores text for intelligent and context-aware conversations.

• View project

This project employs advanced deep learning techniques to forecast NBA player scoring output. By analyzing player statistics from the 2023-2024 season, we developed and compared six neural network architectures, ultimately identifying the BiLSTM model as the most effective predictor.

• View project

Built a scalable model on 458K+ customer records to predict credit card defaults. Applied SHAP for feature reduction (190 → 14), tuned hyperparameters for generalization, and optimized thresholds for risk–reward tradeoffs. Delivered an interpretable, retrainable pipeline aligned with financial compliance standards.

• View project

Built a machine learning pipeline to automate Global Bank’s credit card approval process using 2.4K+ historical applications. Cleaned and encoded 18+ features, handled missing edge cases, and deployed a Random Forest model that reduced manual reviews by 80% and accelerated decision time 5x.

• View project

This project aims to predict the 10-year risk of future coronary heart disease (CHD) in patients using machine learning techniques. By analyzing various health indicators and lifestyle factors, we develop a model that can assist healthcare professionals in identifying high-risk individuals and implementing preventive measures.

• View project

This project implements a robust sentiment analysis model using Natural Language Processing (NLP) and machine learning techniques. The model classifies text into positive or negative sentiments with high accuracy, providing valuable insights for various business applications.

• View project

Developed a Generative Adversarial Network (GAN) model to generate anime-style faces using Python libraries, including TensorFlow, NumPy, and Matplotlib. This project involved designing and training the GAN to produce high-quality, realistic anime faces by implementing advanced techniques and fine-tuning methods to enhance performance and output quality

• View project

Designed and built a fully normalized SQL database (1NF–3NF) from a flat HR dataset, boosting complex query performance by 50%. Developed ER diagrams, applied indexing and partitioning, and ensured referential integrity across tables. Authored advanced SQL queries using CTEs and window functions to analyze attrition trends, pay equity gaps, and education-performance relationships. Optimized the schema for seamless BI tool integration and future scalability.

• View project

This project focuses on developing and advancing a system for real-time monitoring and alerting of potential vehicle accidents.
The repository is divided into two major phases:
1. Initial Project: The foundational work done during the internship at Electronics Corporation of India Limited (ECIL), which involved real-time vehicle data collection and basic alert mechanisms.
2. Advanced Project: An enhanced version with advanced data processing, real-time alert integration, and further optimizations.

• View project

The project is focused on analyzing data collected from Programmable Logic Controllers (PLC) in CNC (Computer Numerical Control) machines. The goal of this project is not only to monitor the operational performance of these machines and detect anomalies but also to provide actionable insights and suggestions for improving machine efficiency and reliability.

• View project