Kalyan Srivastava

Built an OCR→NLP pipeline to extract adverse events from physician reports using regex, NER, and R-based microservices orchestrated with Docker, Jenkins, and Airflow; produced regulatory-grade AE lists and reduced manual review time. Developed automated molecular report generation workflows for sequencing data used in regulatory submissions. Supported development of a radiology-based breast cancer staging model using CNN/BNN architectures and LLM-assisted image interpretation. Assisted in converting SAS code to NextGen analytical tools (DBX PySpark, R/dplyr, SQL) in a Databricks sandbox, and evaluated outcome accuracy with custom comparison tools. Built an RAG-based NLP chatbot using LangChain, transformer models (BioBERT, ClinicalBERT, GPT, BioMistral) and Dialogflow to answer trial development questions from PDFs, reducing literature review time by ~60%. Designed two production-grade Shiny dashboards with custom CSS/JavaScript and NLQ tooling. Implemented CDISC data engineering acr

Provided support to ~120 pharma/clinical users on Cytobank’s ML platform; troubleshooted supervised (logistic regression, XGBoost, neural networks) and unsupervised (t-SNE, UMAP, SOM) workflows on high-dimensional biological data. Collaborated with the development team to build 3 major ML algorithms and 8 novel mini features; performed feasibility assessments and incorporated user feedback for 100 feature enhancements. Led the conversational AI assistant initiative to automate FAQs and streamline user support using Zendesk SDK, Amazon Lex, and GPT. Developed 20+ Python/R scripts leveraging Cytobank APIs for premium users. Managed Agile workflows with Jira/ServiceMax; improved KPIs by 10%. Coordinated with commercial/marketing teams using Salesforce and Google Analytics.

Engineered data pipelines transforming SDTM datasets into ADaM and TLF formats for 30+ FDA submissions. Implemented ETL on AWS and GCP to support data pipelines; tested SDTM data consistency; performed statistical programming (t-tests, ANOVA, z-tests, regression). Applied data wrangling/preprocessing for ML workflows (feature engineering, imputation, scaling, one-hot encoding). Supported ML with SVM, logistic regression, random forest, and XGBoost; enabled 20 concurrent classical/neural models on SageMaker/DataRobot. Developed interactive TLF visualizations in RShiny, PowerBI, and Tableau. Managed a 10-engineer team with Jira/ZenHub/Asana; collaborated with DevOps to maintain APIs and data migration processes.