
Centralized licensing, entitlements, and runtime observability platform engineered with Go, React, Redis, and Kubernetes.
Selected work
A curated selection from 200+ production systems across enterprise AI/ML, agentic architectures and open-source frameworks. Select a project for details.
Showing 65 of 65

Centralized licensing, entitlements, and runtime observability platform engineered with Go, React, Redis, and Kubernetes.

End-to-end AI agent and workflow orchestration system supporting multi-step reasoning, RAG, and human-in-the-loop governance.

Centralized engineering repository and architecture hub for structuring, evaluating, and sharing reusable AI/ML system blueprints.

Production conversational and knowledge-retrieval systems engineered with hybrid RAG, Graph-RAG, semantic reranking, and Langfuse observability.

Advanced data preprocessing library based on Keras preprocessing layers. Standalone or as first layers.

Unified recommendation model combining Collaborative Filtering, Two-Tower Retrieval, and Ranking with dynamic weighting, multi-task learning, and feedback loops.

Modular, production-ready LangGraph application with Gemini LLMs and BigQuery enabling natural-language SQL interactions with isolated memory.

ML-powered fraud detection system analyzing transaction patterns, payment velocities, and behavioral anomalies to prevent financial losses in real-time.

Advanced detection of fraudulent return schemes and parcel claim abuse using graph analytics and behavioral clustering to save €2M+ annually.

Mathematical optimization of warehouse allocation, inventory levels, and distribution routes saving €10M+ annually through constraint programming.

Time-series forecasting of in-store customer demand for staffing optimization and flow management, driving €10M+ in operational savings.

Custom fine-tuned SLM achieving 50% → 85% accuracy improvement on data extraction and RAG tasks with reduced inference costs.

Custom MLOps platforms built from ground up enabling fast deployment, model versioning, A/B testing, and feature stores at scale.

Optimization algorithms for robotic warehouse operations, pallet placement, and picking route efficiency using constraint programming.

AI-powered negotiation training platform with virtual coaching, real-time feedback, and skill metrics for enterprise staff development.

Collaborative HR agents handling employee queries, policy lookups, and automated responses using RAG and document understanding.

ML-powered code analysis detecting secrets, API keys, and security vulnerabilities using semantic pattern matching and graph analysis.

Visual similarity search engine using deep CNN embeddings and vector indexing to identify marketplace products directly from user photos.

Automated deep learning image segmentation pipeline isolating marketplace items onto standardized backgrounds at catalog scale.

Computer vision pipeline extracting dominant product colors and calibrated RGB vectors from user-uploaded photography for catalog enrichment.

Unsupervised deep autoencoder framework for tabular anomaly detection with integrated feature preprocessing and threshold calibration.

Real-time payment and account fraud scoring pipeline combining supervised gradient boosting with unsupervised behavioral anomaly signals.

Dynamic pricing engine for second-hand and unique catalog items using probabilistic Bayesian regression and demand elasticity modeling.

Multilingual NLP entity resolution system matching unstructured listings to canonical catalog references across 6 languages.

Automated time-series anomaly detection and alerting pipeline monitoring marketplace KPIs, data drift, and production model health.

Multi-objective catalog ranking pipeline prioritizing seller curation and inventory merchandising to maximize marketplace GMV.

Omnichannel recommendation and assortment ranking system optimizing physical store inventory allocation and reducing stockouts by 30%.

Comprehensive ML pipeline framework combining simplicity of Python with enterprise MLOps power.

Building blocks—atoms, steps, components—to mix and sequence for customized ML pipelines.

Reusable Keras layers and models for tabular data processing, feature engineering, and neural networks.

Go framework for building AI agent workflows using graph-based execution with performance of Go.

Interactive AI project scoping assistant powered by Small Language Models running locally in the browser via WebGPU.

Customizable QR code generation toolkit with AI-assisted styling, error-correction validation, and export templates.

Desktop Python applications automating laboratory data analysis, plotting, reporting, and centralized database synchronization.

Functional, scalable implementation of Double Machine Learning for Lifetime Value analysis using TensorFlow with comprehensive observability.

Flexible anomaly detection framework combining statistical methods (Z-score, IQR, MAD) with business rules for comprehensive insights.

Advanced conversion detection using TensorFlow autoencoders for customer scoring and behavioral pattern recognition.

In-house MMM using agent-based modeling for marketing efficiency optimization and budget allocation decisions across channels.

TensorFlow-based system optimizing product placement using geographical data, sales patterns, and Haversine-based store clustering.

Predicts product sell-through time using TabTransformer architecture with categorical embeddings and taxonomy features.

Named entity extraction from product descriptions using spaCy and Transformers to enrich catalogue data quality and improve search.

Algorithm transforming tabular data into CNN-compatible images, assigning features to pixels based on similarity for deep learning.

Neural network predicting sales probability based on product attributes, pricing, and additional marketplace inputs.

Keras LSTM model predicting future stock prices using historical data with hyperparameter tuning capabilities.

Product ranking algorithm without customer classification, optimizing display order to increase CTR and conversion rates.

Production FastAPI template with LangGraph integration, Langfuse observability, pgvector memory, rate limiting, and Prometheus metrics.

Drag-and-drop interface for designing Keras preprocessing pipelines with Next.js frontend and FastAPI backend.

React interface for debugging and testing GoLangGraph agents, adapted from LangGraph Studio without production limits.

Comprehensive framework with specialized agents (Router, Summarizer, Memory Compression) and workflow orchestration capabilities.

Interactive calculator measuring and optimizing ROI of Generative AI projects with cost analysis, break-even tracking, and recommendations.

Automated detection of payslip inconsistencies and calculation errors using statistical analysis and business rule validation engine.

ML model predicting DPE energy ratings from household characteristics, implementing official government calculation methodology with 94% accuracy.

RAG-powered chatbot augmenting customer service with legal knowledge base, handling FAQ, common requests, and case routing with human escalation.

Generative AI system producing marketing campaign visuals at scale, with brand consistency, A/B variant generation, and performance-driven optimization.

AI-powered visualization tool rendering furniture and products in customer home spaces for immersive product discovery and purchase confidence.

Ensemble forecasting system combining LSTM, Transformer, and gradient boosting for stock levels and pricing decisions with high accuracy.

R&D on GNN architectures for product relationships, customer networks, and recommendation improvements using PyTorch Geometric.

Probabilistic pricing model using Bayesian inference for demand estimation with confidence intervals and optimal price point recommendations.

R&D on autonomous agent teams for code production, review, and testing with emergent collaborative behavior and self-improvement.

AI agents for automated statistical analysis of A/B tests, significance detection, and recommendation generation.

Causal ML for infrastructure optimization decisions, understanding intervention effects, and counterfactual analysis.

R&D on domain-specific embedding models for improved semantic search, product matching, and recommendation quality.

Building and mentoring high-performance AI/ML teams, establishing best practices, and creating collaborative technical environments.

Strategic AI roadmap development, use-case prioritization, ROI analysis, and technology stack decisions for enterprise AI initiatives.

University teaching, conference talks, workshops, and mentoring programs sharing AI/ML expertise with the broader community.