SHASHIDHAR REDDY NIMMAGARI

Built and own two production AI systems end to end as product owner and AI engineer.

PV-Prognose is a solar energy forecasting product built with LSTM deep learning models using TensorFlow and Keras. Running in production for three years, it delivers automated forecasts three times daily across 42 solar installations nationwide, maintaining a continuous 5-day rolling forecast horizon with over 90,000 forecast cycles completed. Responsible for the full lifecycle including models, Azure MLOps infrastructure, Tableau dashboards, and client delivery.

SalesNavigator is a GenAI-powered market analytics product designed and built from scratch using Python, Azure OpenAI, and Azure Data Factory. It automatically generates market potential scores, positioning recommendations, and marketing text through multi-stage LLM prompt engineering and geospatial analysis. The system runs containerised on Azure Batch with autoscaling and is currently being rolled out to enterprise clients.

Also built and maintain multiple Azure Data Factory pipelines covering data ingestion, model execution, and automated reporting delivery. Built AI agent workflows using n8n combining LLM reasoning with tool use and structured output generation. Developed predictive models for regional sales analysis using geospatial data and built Tableau and Power BI dashboards used by senior stakeholders for strategic planning. Supported and guided colleagues on data science and analytics topics across the team.

Designed and built a full NLP pipeline in Python to automatically detect redundant, incomplete, inconsistent, and missing software requirements in large-scale automotive specification documents. Developed a novel method for incomplete requirement detection using vocabulary-based constraint checking, cosine similarity for redundancy scoring, and TF-IDF vectorisation for feature extraction. Applied K-means clustering with the Elbow method to identify optimal data groupings and evaluated results using a Random Forest classifier achieving 87.5% accuracy, precision, recall, and F1-score. Automated the full data preprocessing workflow using Python, NumPy, Pandas, Scikit-learn, spaCy, and NLTK.

Automated XML data extraction with Python, improving workflow efficiency by 25%. Built dynamic Excel reporting templates with pivot tables and data visualization to support business decision-making.