Suraj Didwania

G

Senior Data Scientist

Gradle Inc

Sep 2022 - Current (3 years 9 months)

MCP Server: Architected and led production of the Model Context Protocol (MCP), a core platform powering LLM-based agentic workflows across IDEs. Delivered enterprise-grade on-prem deployment, authentication, multi-version orchestration, and large-scale data streaming, enabling AI-assisted developer reasoning at scale. Al-Powered Build/Test Failure Summary: Built an LLM-based log understanding system leveraging Sentence-Transformers, ONNX quantization, and pgvector-backed clustering to aggregate and summarize build/test failures, improving observability and providing real-time failure insights in GitHub PRs across enterprise and OSS projects. Predictive Test Selection: Led end-to-end development and deployment of ML system achieving 0.995 r

Senior Data Scientist

Samsung Electronics

Jan 2022 - Sep 2022 (9 months)

Semiconductor Yield Optimization: Led cross-functional team developing ensemble models for Failed Chip Count (FCC) prediction, resulting in 9% defect reduction and $1.5M savings. NLP for Manufacturing Analytics: Implemented transformer-based system using BERT and Hugging Face libraries to analyze semiconductor process data, fine-tuning models for automated root cause analysis reports. Multimodal ML Systems: Designed hybrid models combining sensor data, process parameter, textual logs using NLTK & data pipeline processing terabytes of manufacturing data using Spark, SQL, AWS for real-time model inference.

Data Scientist

Access TeleCare

Jan 2019 - Jan 2022 (3 years 1 month)

Patient Demand Forecasting: Led team to develop Bayesian deep prediction model using automatic feature extractor with LSTM Auto-Encoder, improving prediction accuracy by 40% & enabling proactive resource allocation. Time Series & Reinforcement Learning: Built ML pipeline (SARIMAX + LSTM) achieving 10% SMAPE and 20% SLA improvement and implemented Multi-Armed Bandit algorithms for optimizing telemedicine state allocation. Clinical ML Models & NLP: Created hierarchical multi-modal classification system (84% accuracy, 0.92 precision) for patient bleed risk prediction, achieving and applied NLP for medical records insight extraction.

J

Machine Learning Engineer

JCPenney & Yantriks LLC

May 2017 - Dec 2018 (1 year 8 months)

Nginx Logs on E-commerce: Anomaly Detection for - Identifying Operational Issues on the Website. Time Series Forecasting: Implemented multi-dimensional LSTM networks for large-scale time series data processing using TensorFlow and Bayesian methods such as MCMC achieved an SMAPE of 12%. Digital Advertising Analysis: Implemented ML classifier models (Random Forest, XGBoost, Logistic Regression) for Control & Treatment set if digital advertisements would impact customers, Achieved precision of 0.80.

About

Detail

Timeline

Résumé