Projects

A selection of hands-on projects focused on Machine Learning, Deep Learning, and applied AI. Each project emphasizes problem formulation, modeling, and practical implementation. Source code is available on GitHub.

LLM Fine-Tuning with RAG for Investment Guidance

Developed a domain-specific investing assistant by fine-tuning the LLaMA 3.2 (3B) model on finance and investment datasets. Integrated a Retrieval-Augmented Generation (RAG) pipeline that dynamically pulls information from investment books, company annual reports, and relevant web articles to provide context-aware, data-driven investment guidance.

LLMLangChainVector DBRAG

View Code

Portrait Image to Sketch Translation with Pix2Pix GAN

This project implements a Pix2Pix conditional GAN with a U-Net generator and PatchGAN discriminator to transform facial photos into sketches. Trained on the FS2K dataset, it produces high-quality, structurally accurate sketches through combined adversarial and L1 loss.

TensorFlowPix2Pix GANU-Net

View Code

Density-Aware Dynamic Cluster Splitting Algorithm

Designed a density-aware post-processing algorithm that automatically detects and splits erroneously merged k-means clusters. The method recursively refines clusters using spatial density analysis, removing dependency on correct k initialization and improving robustness for complex, overlapping data distributions.

NumpyPandasScikit-Learn

View Code

Smart EV Charging Scheduler for Self-Driving Vehicles

Designed an automated scheduler for self-driving electric vehicles that optimizes charging assignments based on various data such as battery levels, station availability, and travel distances. Implemented with MiniZinc and constraint programming to ensure efficient, conflict-free scheduling while minimizing total travel distance.

MiniZincCSPOptimization

View Code

Stock Market Trend Prediction using ANFIS

This project applies an Adaptive Neuro-Fuzzy Inference System (ANFIS) to predict short-term stock market trends using technical indicators such as RSI, MACD, and Volume Ratios. By integrating fuzzy logic with neural network learning, the model automatically learns fuzzy rules that capture non-linear patterns in stock price movements.

ANFISPyTorchScikit-learn

View Code

Uber Fare Prediction with Data Cleaning & EDA

Performed rigorous data cleaning and exploratory data analysis to identify key factors influencing Uber fares. Built and refined regression models with regularization. Validated insights and model performance using error analysis and statistical tests.

EDAStatisticsPandasMatplotlib

View Code

Technical Analysis and Predictive Insights on Stock Returns

Developed a Python-based system to compute key technical indicators (SMA, EMA, RSI, MACD, Volume) and analyze stock return distributions under custom conditions. Enables data-driven insights for identifying high-probability bullish scenarios for stock price.

Scikit-learnMatplotlibTA

View Code

Intelligent Crime Solver using Prolog

Built a logic-based AI system using Knowledge Representation & Reasoning to investigate complex murder scenarios. The model evaluates motives, evidence, contradictions, and cover-ups to identify the killer(s), demonstrating advanced logical deduction, structured reasoning, and AI-driven problem solving.

PrologExpert Systems

View Code