Ritesh Ratti
AI & Data Science Director with PhD in Computer Science & Engineering
About Me
AI and ML leader with over 15 years of experience at the forefront of machine learning innovation. I have held leadership positions at renowned global technology companies including EY, Temus, HelloFresh, Delivery Hero, Grab, Pitney Bowes, Samsung, and Oracle. My expertise spans the development of Large Language Models (LLMs), Generative AI, and the implementation of scalable machine learning products across international markets. I have contributed significantly to the advancement of AI and machine learning community through multiple research publications, delivering talks at international conferences and reviewing various research papers for conferences. My commitment to innovation and technical excellence continues to shape the future of AI applications in enterprise environments.
Technology Domains
Specialized expertise across AI, ML, and data technologies
Network Security & Intrusion Detection
PhD research focus on unsupervised learning methods for network intrusion detection, with multiple IEEE publications and conference presentations on NIDS architectures.
Large Language Models & Generative AI
Extensive experience building LLM-powered applications using RAG, PEFT, vLLM, and Agentic AI frameworks for enterprise chatbots and intelligent automation.
Vision Language Models (VLMs)
Delivered VLM-based solutions for defect analysis, and document content extraction.
Computer Vision
Built production systems for image classification, food item detection, watermark identification and object detection.
Natural Language Processing
Designed multilingual text classification , entity extraction , and address parsing systems deployed across multiple countries.
Recommendation & Personalization
Architected recipe recommendation engines, collaborative filtering systems, and Learning-to-Rank models for personalized recommendations.
Generative AI for Content Creation
Pioneered Stable Diffusion-based solutions for personalized marketing assets, customized taglines, and dynamic banner image generation.
Voice & Speech Technologies
Led Voice-to-Voice communication Engine using cloud and OpenAI models
Graph Analytics
Engineered Community Detection using graph analytics for customer segmentation.
Data Deduplication
Designed Smart Data Quality solutions using active learning with clustering and classification for advanced deduplication in enterprise data systems.
Database Security
Developed database security product components like log collectors, frameworks, and command-line utilities for enterprise security.
Distributed Systems
Architected distributed event processing frameworks using Storm and Kafka, Large scale ML model development using Spark and Kubernetes.
Experience
15+ years across global tech companies
EY
Director - AI and Data Science
Data x AI Technology Centre of Excellence
- Leading cross-functional teams comprising Data Scientists, AI Engineers, and Developers to build software solutions
- Driving strategic team performance by implementing quarterly OKRs, prioritizing sprints, and developing project roadmaps while maintaining stakeholder-aligned backlogs
- Co-Lead for EY AI Platform powering Agentic AI solutions in Chatbot, Voice communication, and Predictive analytics domain
- Leading AI & ML projects in Gen AI, Agentic AI and ML space with consistent feedback delivery for AI project execution
Temus
Temasek owned Digital Transformation Company
Manager - AI and Data Science
AI x Data Centre of Excellence
- Led cross-functional AI and Data engineering team providing technical architecture expertise, code review support, and development mentorship
- Delivered Voice-to-Voice Sales Representative Engine using Azure STT, TTS, and OpenAI-4o, reducing onboarding time by 20% for 50+ concurrent users on Kubernetes
- Spearheaded VLM models for defect analysis and website migration error detection, improving defect identification accuracy by 15%
- Led Document Content Extraction system using VLM models to extract text, images, charts, and tables with 95% accuracy
Click to see 2 more...
HelloFresh
Data Science Lead
Personalization Team
- Technical Leader for personalization team, directing data scientists and ML engineers for end-to-end delivery of ML-based personalization products
- Spearheaded Recipe Recommendation and Recipe Ranking solutions using Random Forest for US and Canadian markets achieving 40% average recall@k
- Engineered Cold-start solution using vector database for similarity search and LLM-based Recipe recommendation for new customers
- Pioneered Generative AI solutions using Stable Diffusion for personalized marketing assets including customized taglines and banner images
Click to see 2 more...
Delivery Hero
Senior Data Scientist
Food Data Science Team
- Directed team of 3 data scientists, overseeing project management for food data science division with close regional team collaboration
- Designed multilingual Distil BERT for Food item categorization across 9 EU and 5 APAC countries achieving 80% weighted F1-score
- Delivered Multimodal dish classification using BERT + MobileNet with early fusion techniques, achieving 4-5% performance improvement
- Engineered Community Detection using graph analytics for customer segmentation, resulting in 2% increase in net revenue across Singapore and Dubai
- Architected Food Image classification using MobileNetV2 and EfficientNet for 500 food classes with 75% accuracy, plus Siamese networks for placeholder detection
Click to see 3 more...
Grab
Senior Data Scientist
Data Science Team
- Lead for data science solution development, mentoring team members and collaborating with managers to establish roadmaps for data science initiatives
- Spearheaded computer vision projects including Image Correction and Mart Item Tagging, leading Eater-side recommendation initiatives on big data infrastructure
- Implemented Food Mart Item categorization using Distil BERT for multi-label, multi-class predictions across 6 countries achieving 70% accuracy
- Engineered Food-Nonfood detector (75% accuracy) using MobileNetV2 and Multi-watermark detection (70% accuracy) using ResNet50, enhanced with YOLO-v3
- Developed Up-selling recommendations using FP Growth on Spark MLLib improving AOV by 5%, and Learning to Rank model improving NDCG by 1-2%
- Built time-series models for Food preparation time estimation using statistical analysis of sequential button-press data
Click to see 4 more...
Pitney Bowes
Senior Advisory Software Engineer - ML
Customer Infomrmation Management Team
- Technical Lead for NLP initiatives, spearheading architecture development and transforming project requirements into actionable development stories
- Designed Smart Data Quality solution using active learning with clustering and classification for advanced deduplication
- Developed Address Parser using MLP + Word2Vec achieving 75% accuracy for UK and Germany, with LSTM POC improving metrics by 2-4%
- Architected Entity Extractor using CRF models for CoNLL-2003 and MASC datasets, plus Text Categorization using SVM and MaxEnt
- Published research at ICSC 2018 (Los Angeles) on address parsing and IJCAI SML 2017 (Melbourne) on bank wire entity extraction
Click to see 3 more...
Wynk Limited
Senior Software Engineer - ML
Wynk Recommendation Team
- Led ML development for recommendation systems, contributing to Songs recommendation engine using Apache Storm-Trident aggregation
- Designed dynamic My Favorites feature analyzing user activity to calculate interest scores across Hadoop with Hive-generated analytics
- Implemented Google Now integration for Context-specific cards based on user activity patterns with automated album release notifications
- Architected distributed asynchronous Event processing framework using Storm and Kafka, achieving 500% improvement in transaction processing speed
Click to see 2 more...
Samsung Electronics
Technical Lead - ML
Advanced Technology Group
- ML Lead for Advanced text analytics using tf-IDF, topic modeling, and entity extraction on Hadoop with Elasticsearch indexing
- Designed Topic Extraction using LDA modeling with Question/Answer scoring using extracted topic probability distributions
- Architected map-reduce workflow for Textual Feature Extraction with Apache Oozie automation at Samsung HQ, Korea
- Developed CaaS infrastructure and camera use cases for real-time Contextual configuration suggestions
- Built configuration suggestion system using K-Means clustering with POS Tagging for voice-based command processing
- Engineered optimized caching scheme using Memcached improving object access performance by 20%, authored patent for Cross Region Caching
Click to see 4 more...
Oracle
Member Technical Staff
Database Security Development Team
- Software Developer for database security products, creating Log collectors and CLI tools for Oracle Audit Vault
- Designed and developed OSAUD, DBAUD log collectors and CSDK with push-based event model for AV11g platform
- Implemented logging schemes and Information Lifecycle Management systems for collector metrics
- Developed AVCLI command line utility with console interface including Firewall Management and Administration capabilities
Click to see 2 more...
Technical Skills
Expertise across the AI/ML stack
Programming Languages
Generative AI Technologies
Machine Learning Libraries
Data Processing Tools
Data Storage Systems
Cloud Technologies
Monitoring & Scheduling
Publications
13+ research papers in ML & Network Security
Journal of Information Security and Applications
Novel multiview approach to network intrusion detection combining multiple perspectives for improved threat identification.
IEEE Access
Leveraging prompt engineering techniques with LLMs for network intrusion detection and classification.
IEEE TrustCom 2023
Unsupervised learning approach for network intrusion detection with protocol-specific feature engineering.
IEEE TrustCom-EUC 2023
Deep learning architecture using LSTM autoencoders with temporal awareness for network anomaly detection.
IEEE ANTS 2021
Real-time network attack detection system utilizing statistical feature extraction and machine learning.
IEEE ICCCNT 2020
Scalable intrusion detection approach using entropy-based discretization for efficient network traffic analysis.
IEEE ICSC 2018
Neural network architecture for parsing and understanding geographical address components across formats.
IJCAI SML Workshop 2017
Deep learning approach for extracting named entities and semantic information from financial wire transfer text.
IEEE TrustCom 2012
Active monitoring system for real-time detection of ICMP-based network attacks and vulnerabilities.
International Conference on Computer Science and Information Technology
Host-based detection mechanism for identifying and mitigating ARP spoofing and related LAN attacks.
ISA Transactions
Discrete event systems approach for modeling and detecting various LAN-specific attack patterns.
International Conference on Advanced Computer Science and Information Technology
Active intrusion detection framework specifically designed for detecting attacks in local area networks.
IEEE Mediterranean Conference on Control and Automation (MED)
Discrete Event Systems methodology applied to ARP spoofing attack detection in network environments.
Blog
Technical insights on AI, ML, and Data Science
Moving Forward with AI
AI advancements in logistics and how it is transforming the way we work
Building Chatbot using LLM based Retrieval Augmented Generation Method
Learn how to develop intelligent chatbots using Large Language Models and RAG architecture to utilize knowledge bases for answering complex queries.
An Introduction to Multimodal Machine Learning
Explore computer algorithms that learn from multiple data modalities including text, images, audio, and behavioral data for enhanced AI systems.
Learning to Rank Algorithms
Deep dive into machine learning methods that solve ranking problems using supervised learning to discover optimal ordering for item lists.
Evolution of Recommender Systems
Trace the journey from traditional to modern deep learning-based recommender systems using MLP, LSTM, and attention-based models.
Custom Text Data Augmentation using NLPAug, SentencePiece, and Tensorflow
Techniques for data augmentation to provide varied training data, reduce overfitting, and produce more generalized machine learning models.
Batch Normalization Demystified
Understanding Batch Normalization as a technique for training deep neural networks by standardizing inputs to each mini-batch layer.
Speaking & Conferences
17+ international presentations across the globe
Faculty Development Program, LBRCE
Next Generation AI Architectures for Multimedia Data Analytics
Indian Institute of Technology
Evolution of Recommender Systems
Hasso Plattner Institute
Deep learning based methods for Network Intrusion detection systems
TrustCom-EUC 2023
Network based intrusion detection using Time aware LSTM Autoencoder
TrustCom 2023
Protocol aware unsupervised network intrusion detection system
Delivery Hero Global Tech Summit
Dish Catalog Optimization
Big Data Innovation Conference - PGS
Developing Transformers-based models for Text Analytics
Global IT Security Summit - PGS
Network based attack detection using Machine Learning
IEEE ANTS 2021
Online Network Attack Detection using Statistical Features
AICTE Faculty Development Program, LBRCE
Data Science and its Applications
Faculty Development Program, LBRCE
Applied Data Science: Tools and Techniques
ICCCNT 2020, IIT Kharagpur
Towards implementing fast and scalable Network Intrusion Detection System using Entropy based Discretization Technique
Grab Hackathon
Feature Engineering and Model Building
IJCAI Workshop on Semantic Machine Learning
Semantic extraction of Named Entities from Bank Wire text
International Conference on Advanced Communication and Networking
An active intrusion detection system for LAN specific attacks
Certificates
Global IT Security Summit Speaker
Awards & Education
Education
PhD in Computer Science and Engineering with specialization in AI & Data Science
Indian Institute of Technology, Guwahati
July 2015 - June 2024 | Guwahati, India
Application specific Network Intrusion Detection System (An unsupervised perspective)
M.Tech in Computer Science and Engineering
Indian Institute of Technology, Guwahati
July 2008 - June 2010 | Guwahati, India
Active Detection mechanisms for attacks in Autonomous Systems
CPI: 9.1/10Awards & Recognition
Samsung Employee of the Month
Samsung Research
2014Recognition for outstanding contributions to text analytics solutions
PB Star Award
Pitney Bowes
2016Excellence in NLP and machine learning initiatives
Spectrum CIM Hackathon Winner
Pitney Bowes
2017First place in company-wide hackathon for innovative ML solution
Patent
Method and system for updating a local cache database
Patent for innovative caching mechanism in distributed systems
Professional Activities
• IEEE Technical Program Committee Member (ANTS 2024)
• IEEE Conference Reviewer (ANTS, GCON)
• Active Speaker at International Tech Conferences
Get In Touch
Discuss AI opportunities, mentorship, or collaboration
Connect With Me
Always interested in discussing AI/ML projects, research collaborations, or speaking opportunities.
Schedule a Meeting
Mentorship & Consultation
© 2026 Ritesh Ratti