Traple Kumar

2+ years of work experience

Jul 20242 y 4 moPresent

Bengaluru, Karnataka, India

• Built and enhanced 5+ production GenAI modules across RAG, Text2SQL, AI agents and Vertex AI solutions. • Improved agentic RAG system performance and reduced API response time by 82% using multithreading in production. • Built a GenAI campaign workflow for Mondelez with cross-functional teams to enable targeted personalization. Contributed to an estimated 8-10% ROI improvement and $181M-$241M in projected annual revenue. • Evaluated RAG responses using context precision, faithfulness, and response relevancy to assess quality. • Improved an LLM-powered Text2SQL engine by optimizing generated queries for target cohort data requirements. • Built async FastAPI endpoints using OOP and multithreading, with Pytest to improve service reliability. • Containerized AI workloads with Docker and deployed them on Kubernetes, Vertex AI Registry, and GCP Logging.

May 20232 moJul 2023

Gurugram

• Led a 5-member team to build a scalable NLP search platform for diverse product datasets and use cases. Reduced development cycles by 95% while improving search relevance and user engagement across product data. • Integrated Elasticsearch with the RASA DIET Classifier to enable scalable product search and filtering. • Used BERT embeddings and GPT-3 to enable semantic search and automate classifier training workflows.

Apr 20224 moAug 2022

Remote

• Built a real-time Indian currency classifier using transfer learning with Inception V3 and augmentation. • Achieved 95.61% precision and 94.02% validation accuracy using data augmentation and model optimization. • Integrated voice predictions to support visually challenged users in real-time currency identification.