Nov 2024 – Jan 2026 · 1 yr 2 mos · Singapore, Singapore · On-site

• Leading a cross‑functional AI and Data engineering team while providing technical architecture expertise, code review support, and development mentorship to ensure successful ML product delivery.
• Driving strategic team performance by implementing quarterly OKRs, prioritizing sprints, and developing project roadmaps, while maintaining stakeholder‑aligned backlogs and delivering consistent feedback for AI project execution.
• Managed and delivered an innovative Voice‑to‑Voice Sales Representative Engine leveraging Azure STT, TTS, and OpenAI‑4o model, reducing onboarding time by 20% for 50+ concurrent users on kubernetes based scalable infrastructure, and fine‑tuned OpenAI LLM models to meet specific business requirements.
• Spearheaded the development of innovative solution leveraging VLM models for comprehensive defect analysis, and website migration error detection that helps to reduce quality assurance costs, and significantly improved defect identification accuracy by 15%
• Led end‑to‑end development of Document Content Extraction system utilizing VLM models.

Jan 2024 – Sep 2024 · 8 mos · Berlin, Germany

• Worked as a Technical Leader for personalization team initiatives, collaborating with cross‑functional teams while directing a specialized group of data scientists and machine learning engineers to ensure end‑to‑end delivery of ML‑based personalization products and solutions.
• Spearheaded the development of advanced recommendation engine data science projects including Preselected Recipe Recommendation and Recipe Ranking solutions, designing and implementing a Random Forest‑based model for US and Canadian markets that achieved an impressive 40% average recall@k performance metric.
• Engineered an innovative cold‑start solution utilizing vector database technology for similarity search functionality, while simultaneously developing a POC using LLM based recipe recommendation engine designed to deliver relevant culinary suggestions to new customers.
• Pioneered development of Generative AI solutions for personalized marketing assets, leveraging Stable Diffusion model to create customized taglines and banner images based on customer order history, significantly enhancing marketing effectiveness.

Jan 2022 – Dec 2023 · 1 yr 11 mos · Berlin, Germany

• Directed a specialized team of 3 data scientists and oversaw project management for the food data science division, ensuring end‑to‑end delivery of food science initiatives while maintaining close collaboration with regional teams to successfully implement models, and ML solutions.
• Orchestrated quarterly OKR development in collaboration with project management, establishing comprehensive roadmaps for data science initiatives to ensure successful technical delivery and strategic alignment of projects.
• Designed and implemented a multilingual Distil BERT based model for food item categorization across 9 EU and 5 APAC countries utilizing transformer architecture, achieving an 80% weighted F1‑score, while leveraging these predicted attributes to develop a vendor tagging system based on customer order history.
• Delivered an advanced multimodal dish classification system using early fusion techniques in TensorFlow that seamlessly integrated text and image data through BERT and MobileNet architectures, delivering a 4‑5% performance improvement across key metrics.
• Engineered a sophisticated Community Detection system utilizing graph analytics for targeted customer segmentation based on predicted Dish Attributes, resulting in a 2% increase in average net revenue across Singapore and Dubai markets.
• Architected and deployed sophisticated Food Image classification systems leveraging MobileNetV2 and EfficientNet architectures to accurately identify 500 distinct food classes with 75% accuracy. Also developed Placeholder image detection using Siamese networks with contrastive and triplet loss functions.

Nov 2018 – Nov 2021 · 3 yrs · Singapore

• Worked as a Lead for data science solution development, actively contributing to product development while mentoring team members throughout the development lifecycle and collaborating with managers and product owners to establish comprehensive roadmaps for data science initiatives.
• Spearheaded development of computer vision data science projects including Image Correction and Mart Item Tagging, while simultaneously leading Eater‑side recommendation initiatives leveraging advanced data science algorithms on big data infrastructure.
• Designed and implemented a Food Mart item categorization system utilizing Distil BERT architecture to predict both category and subcategory classifications across 6 countries, leveraging transformers‑based multi‑label, multi‑class models that achieved 70% average accuracy.
• Engineered comprehensive image classification systems including a Food‑Nonfood detector leveraging transfer learning and MobileNetV2 architecture achieving 75% accuracy, alongside a multi‑watermark detection system using ResNet50 with 70% accuracy, further enhanced through YOLO‑v3 object detection implementation that improved performance metrics by an additional 2‑4%.
• Developed Up‑selling items recommendation system using FP Growth Algorithm on Spark MLLib for millions of customers that improve the average order value by 5% and also developed trending Recommendation Learning to Rank model to identify the geo‑hash specific ranks for distance / popularity score / fulfillment rate etc. by implementing logistic regression based model that improved NDCG score by 1‑2 %.
• Developed sophisticated time‑series models to estimate food preparation time using statistical analysis of sequential button‑press data from application interfaces at various preparation stages, enabling more accurate delivery predictions.

Apr 2016 – Oct 2018 · 2 yrs 6 mos · Noida Area, India

• Worked as a Technical Lead for Natural Language Processing (NLP) initiatives, spearheading architecture development while collaborating with stakeholders to transform project requirements into actionable development stories.
• Designed and developed a Smart Data Quality solution utilizing active learning techniques anchored in clustering and classification methodologies, while conducting extensive research on various clustering approaches to effectively group similar records for advanced deduplication.
• Spearheaded development of Address Parser system utilizing Multi‑Layer Perceptron architecture with Word2Vec vectorization for UK and Germany Global Address data, achieving 75% accuracy, while simultaneously developing a proof‑of‑concept using LSTM‑based sequence prediction model with learned embeddings that further enhanced performance metrics by 2‑4%.
• Architected and implemented an Entity Extractor module for identifying person, location, organization, address, and phone number entities utilizing Conditional Random Field models trained on CoNLL‑2003 and MASC datasets, while developing Text Categorization functionality for custom data using SVM and MaxEnt models, and creating entity extraction solutions for bank wire text processing in financial sector.
• Published research paper on address parsing methodologies at ICSC 2018 at Los Angeles, U.S. & published research paper on bank Wire entity extraction approach at IJCAI SML 2017 at Melbourne, Australia.

Feb 2015 – Apr 2016 · 1 yr 2 mos · Gurgaon, India

• Led machine learning development for recommendation systems, contributing to early‑stage feature development of a songs recommendation engine utilizing Apache Storm‑trident based aggregation functionality.
• Designed and engineered a dynamic My Favorites feature that analyzes user app activity to calculate interest scores across Hadoop infrastructure, enabling personalized content prioritization based on sophisticated Hive‑generated analytics reports.
• Implemented Google Now integration to generate context‑specific cards based on user activity patterns, while developing automated new album release notifications for the entire user base.
• Architected and developed a distributed asynchronous event processing framework leveraging Storm and Kafka technologies, resulting in a 500% improvement in transaction processing speed.

Dec 2012 – Jan 2015 · 2 yrs 1 mo · Bangalore

• Served as a Machine Learning Lead for advanced text analytics solutions, actively driving sophisticated feature development utilizing text mining algorithms including tf‑IDF, topic modeling, and entity extraction deployed on Hadoop infrastructure with Elasticsearch indexing for optimized information retrieval.
• Designed and implemented a Topic Extraction system for unstructured data files utilizing Latent Dirichlet Allocation (LDA) modeling, while developing a sophisticated Question/Answer scoring methodology leveraging extracted topic probability distributions.
• Architected a map‑reduce based workflow for Feature Extraction processes and implemented full automation of offline processing pipelines using Apache Oozie at Samsung Headquarters, Korea.
• Designed and Developed CaaS (Configuration as a service) infrastructure components and implemented camera use cases for providing real‑time contextual configuration suggestions to enhance user experience.
• Developed an intelligent configuration suggestion system using K‑Means clustering algorithms, while implementing POS (Part‑of‑Speech) Tagging functionality to accurately identify and process actions from voice‑based commands.
• Engineered an optimized caching scheme utilizing Memcached technology that enhanced object access performance by 20%, and authored a patent for innovative Cross Region Caching methodology

Jul 2010 – Nov 2012 · 2 yrs 4 mos · Bangalore

• Served as Software Developer specializing in database security product development, creating multiple database log collectors and Command Line Interface (CLI) tools for streamlined product administration, while troubleshooting and resolving issues across various releases of Oracle Audit Vault.
• Designed and developed OSAUD, DBAUD log collectors and CSDK aligned with new push‑based event model architecture, successfully implementing these collectors for the AV11g platform.
• Implemented comprehensive logging schemes and Information Lifecycle Management systems for collector metrics, while designing and developing a command line utility (AVCLI) that provides a user‑friendly console interface for interaction and processing, including integrated Firewall Management and Administration capabilities.
• Modified and optimized Makefiles and plug‑in structures while resolving critical issues across multiple AVCLI releases to ensure stable product functionality

Jul 2008 – Jun 2010 · 1 yr 11 mos

• Active Detection mechanisms for attacks in Autonomous Systems
• Description: The work involves detection of various attacks like ARP spoofing, ICMP related attacks using active detection technique. Various attacks like man in the middle, Denial of service due to ARP spoofing are detected and mitigated by the proposed technique. In second phase of the project , Routing related attacks are implemented and detection mechanism is proposed to detect the attacks like , ICMP redirection , Router Discovery spoofing, Connection Reset attacks etc. The project is sponsored by D.I.T. (Department of Information Technology), India.
• Technologies: C , MYSQL , Netflow , SNMP, Softwares like Ettercap, Wireshark, Nmap.