Surendhar Chowdary

● Designed and deployed end-to-end Generative AI systems using LLMs and RAG pipelines, improving response accuracy by 30%+ across enterprise applications. ● Built scalable RAG pipelines using LangChain, Hugging Face, and FAISS/Pinecone for context-aware Q&A with reduced hallucinations. ● Developed and optimized transformer models (BERT, GPT variants) using PyTorch on GPU, reducing inference latency by 25%. ● Implemented embedding pipelines using OpenAI and Hugging Face for semantic search and recommendation systems on large datasets. ● Designed data ingestion and preprocessing pipelines for structured/unstructured data using Python, Pandas, and BigQuery (GCP). ● Built and deployed REST APIs with FastAPI to serve ML/LLM models for real-time inference and scalable integration. ● Leveraged GCP (Vertex AI, BigQuery, Cloud Storage, Compute Engine) for training, deployment, and large-scale processing. ● Containerized applications with Docker and deployed on Kubernetes (GKE) for high availability and fault tolerance. ● Implemented MLOps pipelines using MLflow and Airflow for experiment tracking, versioning, and workflow automation. ● Designed real-time pipelines using Kafka for streaming data and near real-time predictions. ● Performed LLM evaluation using TruLens and OpenAI Evals, improving reliability and reducing hallucinations. ● Applied prompt engineering and fine-tuning (LoRA/PEFT) for domain-specific LLM customization. ● Monitored models using logging, alerting, and drift detection to ensure consistent production performance. ● Collaborated with cross-functional teams to design AI architectures and deliver scalable ML solutions. ● Built NLP pipelines for semantic search, classification, and entity extraction, improving RAG performance.

● Developed and deployed ML models for classification, regression, and clustering using scikit-learn, XGBoost, and LightGBM, improving decision accuracy by 20–25%. ● Performed data cleaning, preprocessing, and feature engineering on large datasets using Pandas and NumPy, improving model performance and stability. ● Built predictive models (Logistic Regression, Random Forest, Decision Trees, Gradient Boosting) for churn prediction and sales forecasting. ● Applied feature engineering (encoding, scaling, outlier treatment), improving accuracy by 15–20%. ● Conducted EDA using Matplotlib and Seaborn to identify patterns, trends, and inconsistencies. ● Implemented evaluation metrics (cross-validation, ROC-AUC, precision, recall, F1-score) for reliable model performance. ● Performed hyperparameter tuning using Grid Search and Random Search to optimize models. ● Wrote SQL queries for data extraction, transformation, and aggregation for ML pipelines. ● Worked on NLP tasks (text classification, sentiment analysis, keyword extraction) using NLTK and scikit-learn. ● Built text preprocessing pipelines (tokenization, stopword removal, stemming, TF-IDF/vectorization). ● Applied deep learning models (LSTM, embeddings using TensorFlow/Keras) for NLP and sequence tasks. ● Implemented clustering (K-Means, Hierarchical) for customer segmentation and behavioral analysis. ● Built deep learning models (CNNs, LSTMs) for image classification and time-series forecasting. ● Developed lightweight ML APIs using Flask for integration into applications. ● Automated data processing and training workflows using Python, reducing manual effort. ● Collaborated with cross-functional teams to deliver data-driven ML solutions. ● Maintained version control using Git for collaboration and code management.