Rohit Kumar

About Rohit Kumar

I’m an AI/ML Engineer passionate about building and scaling intelligent systems that power next-generation digital experiences. At American Express, I design and productionize machine learning models and pipelines that deliver personalized recommendations, fraud detection, and real-time decisioning for millions of customers worldwide.My expertise spans the full lifecycle of ML in production — from data ingestion and feature engineering (Spark, Kafka, Bigtable) to model training (XGBoost, CatBoost, YOLO, Transformers) and deployment at scale via microservices (Spring Boot, REST APIs, GCP, Kubernetes).Some of my key contributions include: • Personalization systems driving $18M+ in incremental revenue through NLP-driven feature pipelines. • Model-as-a-Service platforms serving >10M daily requests with sub-100ms latency. • Enterprise AI assistants integrating LLaMA, LangChain RAG, and vector databases to deliver Stack Overflow–style developer knowledge summaries — reducing resolution times by 40%. • Computer vision pipelines (YOLO-v5) with real-time object detection at 30 FPS. • Early experimentation with Generative AI, applying LLMs for semantic retrieval, abstractive summarization, and conversational assistants for enterprise productivity.I hold a B.Tech from IIT Guwahati (Institute Silver Medalist) and have been recognized with the American Express Inventor Award for my contributions to patentable AI solutions.I’m driven by the challenge of translating cutting-edge ML and Generative AI research into scalable, production-ready systems that create measurable business impact. Always open to collaborating on AI-driven personalization, recommendation systems, and GenAI-powered enterprise applications.

Experience

1. Ai Engineer 3

   American Express

   Jul 2025 — Present

   Engineered an enterprise-grade AI knowledge assistant integrating LLaMA with a pgVector–backed semantic index and a LangChain RAG pipeline to perform low-latency, in-context retrieval over organization-wide Slack data; delivered Stack Overflow–style abstractive summaries of developer queries, achieving ~40% faster resolution times and a ~35% drop in redundant questions- Architected and delivered a scalable, Model-as-a-Service microservice for XGBoost and CatBoost ranking models—combining Spark-based feature pipelines with a Spring Boot service to expose secure RESTful endpoints for real-time scoring (>10 M requests/day at sub-100 ms latency); integrated with GCP Pub/Sub for inference orchestration-Designed and deployed a real-time triggers pipeline to capture customers\' intent through GCP Pub/Sub, integrating beacon clickstream, email, and chat data for content arbitration.

Education

• Indian Institute of Technology, Guwahati

  Bachelor's degree, Mechanical Engineering

  2017 — 2021