Experience | Pradipta Deb

Hornetsecurity is a leading global provider of cloud security, compliance, backup, and security awareness, protecting over 6 billion mailboxes worldwide.

Leading AI research initiatives for email security platform protecting 6B+ mailboxes globally
Developed and deployed LLM-based detection (>95% acc rate) for Business Email Compromise (BEC) attacks
Implemented fine-tuned language models for sophisticated threat detection (spam, phishing, impersonation)
Pioneered integration of Vision LLMs (Pixtral) for webpage scam detection, reducing false positives by 40%
Architected scalable ML pipeline processing billions of daily emails while maintaining GDPR compliance
Designed agent-based architecture for autonomous threat investigation and response
Mentored and established best practices for LLM deployment, monitoring in production

The Messaging, Malware and Mobile Anti-Abuse Working Group (M3AAWG) is where the industry comes together to work against botnets, malware, spam, viruses, DoS attacks and other online exploitation.

Working with international group of 50+ orgs in developing AI/ML standards for cyber abuse prevention
Established framework for responsible LLM deployment in email security, adopted by major industry players
Coordinating cross-industry initiatives for secure and privacy-preserving AI in cybersecurity
Authored technical guidelines for large language model deployment in anti-abuse system
Provided talks and tutorial sessions regarding LLM and its safe uses

Contract.Fit specializes in AI-powered document processing and information extraction solutions for enterprises.

Led development of transformer-based NLP systems for information extraction across 140+ document types
Developed computer vision solutions for detecting embedded objects in documents
Architected end-to-end ML pipeline processing 100K+ daily documents with 98% accuracy with automation
Served as Lead Developer and Tech Analyst for 4 major clients, bridging technical solutions with business requirements
Collaborated with sales/delivery teams to ensure smooth project deployment

Magnet is a research team at INRIA Lille Nord Europe focusing on machine learning methods for mining information networks.

Developed collaborative computation platform for privacy-preserving statistical analysis using U-statistics
Implemented decentralized learning framework with differential privacy, reducing privacy budget by 45%
Created domain-specific language for privacy policy specification in collaboration with PhD researchers
Optimized homomorphic encryption for low-resource devices, reducing computation overhead by 60%

The German Research Center for Artificial Intelligence (DFKI) is a leading research institution in innovative software technologies.

Development of MaryTTS, improving voice quality by 40% through advanced neural synthesis techniques
Selected in Blizzard Challenge 2018 and participated in Blizzard 2017 for speech synthesis
Developed multi-speaker, multilingual speech synthesis system using deep learning for master thesis
Innovated hybrid approach for grapheme-to-phoneme conversion, achieving state-of-the-art accuracy in 5 languages
Created automated testing framework reducing development cycle time by 50%

Tata Consultancy Services is a global leader in IT services and consulting.

Developed and maintained enterprise applications for clients including Johnson & Johnson and ALCOA
Specialized in full-stack development using Java, C++, and Python technologies
Collaborated in cross-functional teams to deliver high-quality software solutions
Implemented automated testing frameworks improving code coverage by 40%

Professional Experience

Research Engineer

Vice Chair, Artificial Intelligence Committee

Senior AI Engineer

Research and Development Engineer

Research Assistant

Software Engineer