Pickup Machine Learning Engineer at Snapp

Built and maintained content-based and LSTM recommender models using PyTorch and ML pipelines, decreasing ride-booking time by 50%. Organized a clustering service over large datasets using Apache Spark with scalable service design, reducing clustering time by 75%. Designed and developed a robust ETL pipeline using Apache Airflow, PostgreSQL, and data quality checks to handle incremental updates across a 30 GB dataset, ensuring data consistency and performance at scale.

Senior Machine Learning Engineer at Snapp

Developed and deployed a scalable Ranker-as-a-Service framework using Seldon Core, Python, PyTorch, PyTorch, Scikit-learn, and microservices architecture for geospatial search systems, leading to performance improvements (e.g., 7%+ in ranking metrics). Optimized processing pipelines with distributed systems and efficient algorithms, achieving 100x reduction in text similarity processing time. Built and maintained a reverse geocoding microservice using Golang with 10k+ requests per second and 99.9% uptime, enabling low-latency, high-availability production performance. Implemented vector-based search using Sent2Vec models and Elastic search, achieving 12% improvement in query handling efficiency. Fine-tuned the DistilBERT model for Named Entity Recognition (NER) over geospatial data using TensorFlow, boosting 30% search intent detection. Maintained a 100 GB Elastic search cluster with high-availability configurations, supporting ~3k real-time data retrieval requests.