Harshit Singh

Experience

• 

  Paperplane

  Aug 2021 - Nov 2021

  Machine Learning Engineer

  • Fine-tuned a CART-based drug recommendation model with Bayesian optimization, improving accuracy by 15%.• Built a graph pipeline (Python, NetworkX) to map symptom-diagnosis-drug relationships, reducing manual analysis time by 20%.• Leveraged parallel processing with Scikit-Learn and JAX's vmap, boosting classification accuracy by 7% on 1M+ patient records.

• 

  Tata Consultancy Services

  Jun 2022 - Sept 2022

  Machine Learning Engineer

  • Benchmarked Random Forest and XGBoost models for trading signal classification, increasing prediction accuracy by 25%.• Developed an end-to-end ETL pipeline using Python, SQL, and AWS (EC2, S3, SageMaker) to process 1M+ daily data points• Optimized LSTM-based time series forecasting with bidirectional LSTMs and attention mechanisms, improving long-range pattern recognition and reducing latency by 15% through batch inference and selective pruning.

• 

  Palm Beach County

  May 2024 - Dec 2024

  Software Engineer

  • Developed scalable ML pipelines for a Retrieval-Augmented Generation (RAG) system, integrating ElasticSearch and Neural Seek toimprove chatbot response accuracy to 96% through better feature engineering and inferencing.• Optimized data workflows by implementing preprocessing pipelines (fine-tuning Granite 13B on IBM Watson) and multi-threadedingestion with Python and Selenium, reducing manual data handling by 80%.• Enhanced data and infrastructure performance by 30% for a platform serving 1.6M+ residents by optimizing SQL queries withB-Tree indexing and partitioning strategies, ensuring real-time data availability for continuous model training and RAG pipelines.• Automated operational workflows (snapshots, backups, and patch updates) using Python scripts and Oracle RMAN, reducing down time by 60% and enabling MLOps best practices for data ingestion and inferencing. Show less