Project Showcase

A selection of projects demonstrating my practical, results-driven approach to solving business challenges with technology.

Pipeline diagram of the AlphaForge 9-stage quantitative equity trading system
Beta

AlphaForge

Institutional-Grade Equity Long-Short Engine

5-model ML ensemble with Ridge stacking, Black-Litterman optimization, and 4 anti-overfitting gates. Factor-neutral construction achieved -2.3% max drawdown through COVID with Sharpe 1.54 — driven primarily by factor-neutral construction and institutional risk management.

Key Highlights

  • Spec-First Engineering at Scale: Complete formal specification defining every pipeline stage, risk constraint, and validation gate before a single line of code — every architectural decision traces back to a formal requirement.
  • 5-Model ML Ensemble: XGBoost, LightGBM, CatBoost, LightGBM-LambdaRank, Ridge stacked via Ridge meta-learner retrained quarterly on expanding out-of-sample predictions.

Technology Stack

LightGBM
Python 3.11
XGBoost
CatBoost
Screenshot of the Clinlux institutional knowledge platform for healthcare
Beta

Clinlux

Giving Time Back to Care

A comprehensive platform that transforms how healthcare institutions manage clinical guidelines — from automated literature monitoring and AI-proposed updates to conversational access, expert validation workflows, and personalized learning dashboards.

Key Highlights

  • Conversational RAG with Citations: Streaming chat with hybrid retrieval (vector + keyword search) and full citation transparency, so every answer is traceable to its source guideline with evidence chains.
  • Automated Monitoring & Validation: AI monitors medical literature and proposes guideline changes. Experts review proposals with side-by-side diffs in a structured validation queue before adoption.

Technology Stack

OpenAI GPT-5
React 18
FastAPI
Python
Architecture diagram of the Swapfresh event-driven swap cleanup daemon
Production

Swapfresh

Event-Driven Swap Cleanup Daemon for Linux

A production-grade Linux daemon that intelligently reclaims swap space when safe, using kernel PSI triggers and a safety-first design with zero external dependencies.

Key Highlights

  • Hybrid Reactive/Proactive Architecture: Combines Linux PSI triggers for instant reactive response with periodic proactive checks, ensuring swap is reclaimed both during and after memory pressure events.
  • Safety-First Design: Multiple safety guards including a 2x free RAM requirement, exponential backoff on failures, and graceful degradation ensure the daemon never makes memory pressure worse.

Technology Stack

Python 3.10+
Linux PSI
systemd
Event-Driven
Architecture diagram of the Workflow+ Architect multi-agent code generation system
Production

Workflow+ Architect

Multi-Agent Code Generation System

Workflow+ Architect transforms natural language into production-ready Workflow+ scripts for AlphaCom's ERP. Features parallel agent orchestration, evidence-based validation to prevent hallucinations, and continuous learning from user feedback.

Key Highlights

  • Multi-Agent Orchestration: A dynamic supervisor analyzes requests and spawns specialized workers in parallel—documentation search, example retrieval, and schema inspection—synthesizing results into coherent responses.
  • Intelligent Search Architecture: Dual-layer search combining BM25 full-text and vector embeddings with Reciprocal Rank Fusion. HyDE generates hypothetical documentation for better recall, while tiered escalation and a 1000-entry embedding cache reduce API calls by ~50%.

Technology Stack

OpenAI GPT-5
Blazor
.NET 9.0
C# 12
Conceptual diagram of the GenericAi Plugin for Workflow+
Production

GenericAi Plugin for Workflow+

AI Automation for Enterprise ERP

An AI integration plugin for AlphaCom's Workflow+ automation engine, transforming how their clients interact with their ERP system by bringing powerful, enterprise-grade AI capabilities directly to their workflow.

Key Highlights

  • Multi-Provider Architecture: Engineered to support a unified interface for OpenAI, Anthropic, and Google Gemini APIs.
  • Seamless Enterprise Integration: Integrated deeply into the Workflow+ plugin architecture, complete with licensing and cost-tracking functionality.

Technology Stack

.NET Framework 4.8
C# 9.0
WorkflowPlugin SDK
OpenAI API
Screenshot of the Fanktank.ch homepage
Production

Fanktank.ch Website

This website, my digital home.

A sophisticated multi-agent RAG system with hybrid search, real-time process visualization, and production-grade AI orchestration. Not just a portfolio—a live demonstration of cutting-edge AI engineering.

Key Highlights

  • Multi-Agent Orchestration: 6 specialized agents working in concert: Router, Search Specialist, Fact Extractor, Briefing Officer, and more.
  • Hybrid Search + Reranking: Semantic vectors + BM25 keywords + Cohere reranking for state-of-the-art retrieval accuracy.

Technology Stack

OpenAI API
Next.js
React
TypeScript
Screenshot of textSelector in action on an Android device.
Beta

textSelector

Fixing Mobile Text Selection

Born from personal frustration, this app brings desktop-like text selection and real-time search to Android.

Key Highlights

  • Desktop-like Selection: Intuitive shift-click style text selection on mobile.
  • Real-time Search: Instantly find content within selected text.

Technology Stack

Kotlin
Android SDK
Jetpack Compose
MVVM
Interface mockup of the Labela image captioning platform.
Beta

Labela

AI Image Captioning Platform

A platform for AI-powered image captioning, designed for batch processing with customizable prompt templates and efficient review tools.

Key Highlights

  • Batch Processing: Caption multiple images efficiently in one go.
  • Customizable Templates: Tailor prompts for perfect caption generation.

Technology Stack

OpenAI API
React
Next.js
TailwindCSS
Diagram illustrating RAXPLORER's architecture and components.
Alpha

RAXPLORER

Advanced RAG Platform

An experimental RAG system that pushes beyond simple text search to understand document layout, using hybrid retrieval and flexible LLM integration.

Key Highlights

  • Spatial Awareness: Understands document layout for better context.
  • Hybrid Retrieval: Combines semantic and keyword search methods.

Technology Stack

PyTorch
React
Transformers
Vector DB
Example of HandScript recognizing various handwriting samples.
Alpha

HandScript

Personalized Handwriting Recognition

A system that adapts to individual handwriting using fine-tuning, making handwritten document management more accessible and accurate.

Key Highlights

  • Personal Adaptation: Fine-tunes to recognize your unique handwriting style.
  • Document Management: Tools to organize and search handwritten content.

Technology Stack

PyTorch
React
TrOCR
LoRA
Visualization of EchoQuest's audio processing and analysis workflow.
Alpha

EchoQuest

Interview Analysis System

An end-to-end pipeline for analyzing recorded interviews, featuring audio enhancement, speaker identification, transcription, and AI-powered Q&A.

Key Highlights

  • Audio Enhancement: Improves audio quality for better transcription.
  • Speaker Recognition: Identifies who's speaking throughout the recording.

Technology Stack

PyTorch Audio
Whisper
Pyannote
LangChain
AI Projects, RAG Systems & Automation Case Studies | Fanktank