Every Project Started With a Problem Worth Solving

A production deep learning pipeline performing multi-location spatio-temporal inundation depth forecasting. Ingests multi-modal meteorological and hydrological time-series, fuses spatial and sequential features, and outputs calibrated depth forecasts across a city-scale sensor network — deployed end-to-end on GCP Vertex AI with TFX orchestration.

TensorFlow · TFX · GCP Vertex AI · Keras · Python

A province-scale geospatial deep learning pipeline predicting fire occurrence and spread probability from fused satellite, meteorological, and topographic features. Outputs geospatial risk maps derived from vegetation indices and spatial embeddings — deployed on GCP Vertex AI with full TFX orchestration.

TensorFlow · TFX · GCP Vertex AI · Keras · Python

A deep learning-based risk assessment system for autonomous drone delivery of hazardous materials in urban environments. Trains on operational data to produce dynamic safety assessments informing route planning, regulatory compliance, and incident prevention.

TensorFlow · Keras · ANNs · Python

A fully autonomous research pipeline built on the OpenAI SDK where agents independently control the entire search strategy — topic decomposition, query formulation, source evaluation, and synthesis — without any user intervention beyond the initial query.

OpenAI SDK · Multi-Agent · Gradio

A configurable research pipeline built on the OpenAI SDK where the user defines scope, source preferences, depth, and output format. Agent execution is controlled by code rather than autonomous decision-making — combining the efficiency of agents with human-directed precision.

OpenAI SDK · Workflow AI · Gradio

A six-agent sequential pipeline that treats job applications as an intelligence problem. Autonomously researches the target company, extracts ATS keywords, rewrites the resume for relevance, and produces a tailored cover letter — all reviewed by a QA agent before delivery.

CrewAI · Multi-Agent · Gradio

A production-ready MLOps pipeline for telecom customer churn prediction. Covers the full lifecycle — model training, experiment tracking, automated CI/CD, real-time data drift detection with Evidently AI, and FastAPI serving — built to the standards of a production engineering team.

FastAPI · EvidentlyAI · MLflow · Docker · GitHub Actions

A production-style RAG system that ingests PDF documents, indexes them using hybrid BM25 lexical and FAISS vector search, and serves precise, source-grounded answers via a FastAPI endpoint — built with enterprise document intelligence architecture patterns.

FastAPI · FAISS · BM25/Whoosh · LLMs · PyPDF · Python

A TFX-based ML pipeline orchestrated with Apache Airflow — automating data ingestion, schema validation, feature transformation, model training, and evaluation through modular DAG-based tasks. Runs in Dockerized environments for clean portability across teams.

TFX · Apache Airflow · TFDV · TFT · TFMA

A Python library for automated hyperparameter tuning using Particle Swarm Optimization and Pattern Search. Acts as a drop-in replacement for GridSearchCV, converging faster to optimal configurations with native scikit-learn compatibility.

Python · scikit-learn · PSO · Pattern Search · PyPI

A custom TFX component that fills a gap in the standard ecosystem — computing per-output and global metrics for multi-output TensorFlow models and writing TFMA-compatible JSON artifacts for downstream pipeline integration.

Python · TensorFlow · TFX · TFMA · Pipelines · PyPI

A lightweight Python library that transforms geospatial shapefiles into ML-ready graph networks using NetworkX. Bridges the gap between GIS data formats and graph-based ML workflows by parsing shapefile geometry into structured representations.

Python · NetworkX · Fiona · Shapely · NumPy · PyPI

A binary classification model predicting diabetic retinopathy probability from retinal image features. Uses backward elimination and AIC for variable selection on 1,151 patient records — providing a clinically interpretable probability score for early DR detection.

Clinically interpretable probability score for early diabetic retinopathy detection

R · Binary Logistic Regression · AIC

A binary classification system predicting 10-year coronary heart disease risk from patient vitals and lab results. Benchmarks Random Forests against Mixture Discriminant Analysis across 10 random seeds on 4,080 Framingham Heart Study patients for consistent, reliable predictions.

84.3% classification accuracy

R · MDA · Random Forest

A ConvLSTM deep learning model that combines spatial feature extraction with temporal sequence modeling for automated video classification. Processes raw video frames through an OpenCV pipeline and outputs class predictions for video-heavy operational workflows.

84% accuracy and 97% AUC

Python · TensorFlow · OpenCV