Muhammad Abulmagd

• Collaborated with a 5-person tech team to build and improve Batvoice (Django, Vue.js) platform’s performance, boosting system efficiency by 20–70% across different API endpoints. • Designed and delivered a compliance rules engine with a visual rule builder, eliminating error-prone manual JSON/database configuration and reducing implementation time from days to a few hours. • Developed 40+ data visualizations (Vue.js, Highcharts, REST APIs) that translated raw data into actionable insights, dramatically reducing manual reporting time. • Played a key role in client demos and POCs, directly supporting sales and customer acquisition. • Helped scale Batvoice from MVP to production, analyzing 25k+ calls per day across multiple clients.

• Led end-to-end NLP system development, from data preprocessing and model training to deployment as Python microservices for large-scale emotion detection and sentiment analysis. • Owned the end-to-end lifecycle of French NER models, from data preparation and training to production deployment within the anonymization pipeline, ensuring GDPR-compliant processing at scale. • Co-developed Labelit an in-house annotation tool (Django, Vue.js) for audio & text (transcription, NER, classification, audio region annotation), increasing data collection efficiency • Trained and fine-tuned a domain-specific ASR model (Kaldi) for customer support that achieved a lower Word Error Rate (WER) than leading STT providers (Google, AWS).