Projects & Case Studies

A PowerShell script that provisions Ubuntu 24.04 on WSL 2 with Python, Docker, Zsh (Oh My Zsh), systemd, and more—fully idempotent and user-configurable.

Fisher Information Matrix-guided adaptive LoRA rank allocation — experiment harness for efficient fine-tuning of large language models.

Production-quality Python library for feature engineering, schema validation, and data quality profiling at scale.

Complete orchestration layer for Learning-to-Rank models with pluggable feature stores, backtesting, A/B testing, and FastAPI serving.

Python package for validating ML models before deployment with automated checks for performance, data drift, failure modes, and CI/CD integration.

Lightweight self-hosted model monitoring solution for drift detection, latency tracking, and intelligent alerting.

Multi-Model Cost Optimizer for Agentic Workloads - optimize LLM API costs across multiple models and providers.

Reference implementations and tooling for production LLM deployment with best practices and patterns.

AI Risk Assessment & NIST AI RMF-Aligned Model Card Generator for responsible AI documentation.

Modular ML inference pipeline for real-time recommendation systems with high throughput and low latency.

Vendor-Neutral ML Inference Benchmarking Framework for comparing and evaluating inference performance across platforms.

Toolkit for robust LLM fine-tuning on noisy training data with techniques for handling data quality issues.

A deep learning model for accurate human figure segmentation in images using PyTorch and U-Net architecture.