Recently deployed projects
Showcasing innovative solutions across machine learning, data management, and AI orchestration.
SPLADE Model Trainer
Client: Internal R&D, PortugalThis project implements a machine learning system for SPLADE (SParse Lexical AnD Expansion) model training and evaluation. SPLADE is a neural information retrieval model that creates sparse vector representations of text, allowing for efficient document search and retrieval.PLU Finder
Client: PLU Finder, PortugalPLU Finder is a comprehensive digital platform serving as the ultimate resource for produce identification and information, featuring a database of over 1,500 PLU (Price Look-Up) codes enhanced with detailed nutritional content, storage guidelines, and identification details.Enterprise Vault Server
Client: Cellebrite, IsraelThe Enterprise Vault Server (EVS) for UFED is as a digital vault in the network that enables UFED users to access Premium capabilities. These capabilities include advanced features such as unlocking and extracting data from locked or high-end Android and iOS devices.
The new EVS is an excellent example of state-of-the-art programming where UI/UX meets hardcore Rust cross-platform capabilities, covering both EVS-on-Premium and SaaS platforms.
ArangoDB Hybrid Search
Client: Internal R&D, PortugalThis project implements an advanced document search system using ArangoDB with vector embeddings.
The system supports features like chunking large documents, multilingual processing, hybrid search (combining dense and sparse embeddings), loading documents from various sources, and synchronizing data between ArangoDB instances. It's a comprehensive toolkit for implementing semantic search capabilities backed by vector databases.
WolfGPT
Client: Ariston/Wolf GmbH, GermanyWolfGPT combines multiple LLM engines with advanced intention analysis that anticipates user needs before they're expressed. The system seamlessly integrates specialized tool-calling capabilities and robust data processing frameworks, enabling complex operations across varied datasets. This intelligent orchestration layer creates a unified experience where multiple AI models work in concert to deliver precise, contextually aware responses even for ambiguous queries.TinyCrew
Client: Internal R&D, PortugalTinyCrew is a TypeScript framework that orchestrates multiple AI agents to collaboratively solve complex tasks. It features a robust agent system with specialized AI assistants that have distinct goals and tools, coordinated by a central crew manager that assigns tasks to appropriate agents based on their capabilities. The framework includes a shared memory system for knowledge transfer between agents, extensible tool integration for real-world interactions, comprehensive event monitoring for tracking progress, error handling mechanisms, and detailed logging.Vidifo
Client: Internal R&D, PortugalVisual Digital Forensics solution for law enforcement agencies. Utilizing multiple Large Language Models (LLM) to analyze image content to identify illegal activities, threats, and other issues that can be flagged to law enforcement for a detailed review.Piirex
Client: Cellbrite, IsraelPiirex is a high-performance log sanitization toolkit designed to detect and redact personally identifiable information (PII) from log files. The system uses advanced pattern matching with optimized regular expressions to identify sensitive data like email addresses, IP addresses, usernames, and credentials in various log formats.