Ahmed Mostafa

Ahmed co-founded Pigro.ai in 2018 and served as Chief Technology Officer (CTO), leading both the technical development and AI engineering efforts of the company.

Pigro began in Italy as a chatbot engine for enterprises, built around a proprietary algorithm capable of extracting and understanding information from multiple documents and sources. Over time, the company evolved into a provider of enterprise search and retrieval-augmented generation (RAG) systems, helping organizations access and interact with internal knowledge through natural interfaces, including chatbots.

Pigro collaborated with major enterprises in Italy and Switzerland, including MSC, BNP Paribas, Sky, Amadori, and several Italian municipalities.

Key Responsibilities:

🔹 AI Research & Development Designed and implemented machine learning and deep learning models for a variety of natural language processing (NLP) tasks, including:

Automated Question Answering

Text Classification

Automated Question Generation

Text Summarization

🔹 Technical & Product Strategy Defined and executed the company’s engineering roadmap, ensuring scalable and efficient AI solutions aligned with business goals.

🔹 Business Development & Company Strategy Actively contributed to Pigro’s strategic direction by engaging with clients and investors, identifying market opportunities, and refining the company’s AI-driven value proposition.

🔹 Team Management Built and led a remote engineering team, fostering a collaborative culture focused on innovation, agility, and continuous learning.

Meedan is a nonprofit organization focused on building digital tools and community-driven programs that enhance the accessibility, reliability, and distribution of information, especially where and when it is needed most.

As a Machine Learning Engineer and Researcher at Meedan, he has contributed to several of the organization's open-source projects, including:

Check: A verification platform designed for fact-checking and collaborative investigations. https://github.com/meedan/check

Alegre: A multilingual and multimodal similarity search engine for content analysis. https://github.com/meedan/alegre

Presto: A serving framework for deploying AI models. https://github.com/meedan/presto

He developed machine learning models for text, image, audio, and video similarity, enabling sophisticated cross-modal content matching and verification.

In addition, he co-authored a peer-reviewed research article published in the International Journal of Public Opinion Research: https://academic.oup.com/ijpor/article/36/3/edae032/7709027

His work also includes contributions to natural language processing models, such as:

An embedding language model for Filipino and Tagalog: https://huggingface.co/meedan/paraphrase-filipino-mpnet-base-v2

A binary classifier for detecting whether input text relates to Brazilian elections: https://huggingface.co/meedan/brazilianpolitics

He conducted his master’s thesis in collaboration with Deepset, under the supervision of Prof. Aris Anagnostopoulos (Sapienza University of Rome) and Mr. Timo Möller (Deepset).

🔹 Thesis Title: Applications of Cross-Lingual Language Models in Question-Answering Systems

This research focused on the use of cross-lingual language models for question-answering tasks, particularly in low-resource languages. By leveraging both English and Spanish datasets, the study employed a dual strategy:

✅ Training data available in the target language

✅ Machine-translated datasets to augment training corpora

The proposed approach surpassed the state-of-the-art (SOTA) at the time on multiple evaluation benchmarks, showcasing the effectiveness of multilingual learning techniques in improving QA systems for underrepresented languages.

The work was implemented using Deepset’s FARM library: https://github.com/deepset-ai/FARM