Harsha Singh

I am a Lead Machine Learning Engineer with 7+ years building production-grade ML solutions and retraining systems. I hold the Google Cloud Professional Machine Learning Engineer certification.

At HSBC (May 2021–present) as AVP - Data and Analytics, I led GenAI initiatives producing synthetic call transcripts with Text-to-Speech and generated synthetic fraud patterns that improved anomaly detection and reduced fraud losses by about 2%. I also built a GenAI-driven fraud-investigation automation that delivered efficiency gains equivalent to 12 FTE.

I redesigned inference pipelines to cut model runtime from overnight to 90 minutes, led feature-store framework design while evaluating Vertex AI for scalable reuse, and integrated Google Vision and NLU to improve document validation accuracy by 10%. I automated voice call audits using DeepSpeech, custom NLU, Airflow, and Bash to reduce manual overhead.

Previously at IndiaMART I developed Transformer-based seq2seq models for query correction, BiLSTM-CRF models for NER, and BiLSTM-Attention classifiers for search type and language detection, deployed with TFX and Docker on GCP. I combine expertise in NLP, computer vision, MLOps, PyTorch/TensorFlow and Big Data to deliver scalable, production ML systems that drive measurable business impact.