Pavani Gulla

09/2024 — Present

• Developed a Gen AI-powered in-vehicle voice assistant platform using LLMs, RAG, contextual embeddings, and conversational memory, enabling
• intelligent diagnostics, infotainment control, navigation assistance, and personalized driver interactions via collaboration with engineering teams.
• Built multimodal GenAI pipelines integrating voice inputs, vehicle telemetry, infotainment systems, and driver context using LangChain, RAG
• frameworks, advanced feature engineering, and prompt optimization on automotive datasets including vehicle manuals, service logs, DTC codes,
• and driving records for real-time assistance, troubleshooting, and personalized recommendations, improving response accuracy by 32%.
• Designed foundation model fine-tuning workflows using Llama, Mistral, and GPT architectures with distributed training, Bayesian hyperparameter
• tuning, instruction tuning, contrastive retrieval optimization & synthetic data cohort for conversational intelligence on automotive data repositories.
• Architected edge inference systems using TensorRT, ONNX Runtime, quantized LLM deployments, and NVIDIA Jetson platforms, enabling low-latency
• conversational AI processing within vehicles, reducing cloud dependency and improving inference response time and system efficiency by 35%.
• Established streaming AI infrastructure using AWS IoT Core, Kinesis, and Flink for real-time vehicle telemetry ingestion, enabling contextual
• awareness, proactive driver assistance, personalized in-car experiences, and intelligent event monitoring, improving data processing output by 44%.
• Implemented agentic AI and GraphRAG systems using LangChain, LangGraph, MCP orchestration, vector databases, and knowledge graphs,
• achieving 0.92 F1-score and 94% precision through rigorous evaluation, A/B testing, shadow deployments, and scalable CI/CD practices.

06/2022 — 08/2023

• Developed automotive anomaly detection solutions for connected mobility and ADAS ecosystems, analyzing real-time vehicle telemetry, sensor,
• diagnostics, and operational event streams, improving incident detection speed by 28% while strengthening fleet monitoring capabilities.
• Designed scalable ETL pipelines using Azure Data Factory, PySpark, and SQL for vehicle telemetry, diagnostics, ADAS, and connected mobility data,
• processing millions of daily records supporting analytics, reporting, regulatory compliance, and automation workflows.
• Built and optimized ML models including XGBoost, LightGBM, and Autoencoders for predictive maintenance, vehicle health monitoring, battery
• performance analysis, and anomaly detection, improving predictive accuracy by 32% across production automotive environments.
• Applied NLP techniques using BERT and TF-IDF Logistic Regression on diagnostic logs, warranty claims, service reports, and customer feedback,
• identifying recurring vehicle performance issues and improving issue prioritization efficiency by 42% organization-wide.
• Containerized ML solutions using Docker and deployed them on Azure Kubernetes Service for real-time inference supporting predictive
• maintenance, fault detection, driver assistance analytics, and connected vehicle intelligence, reducing inference latency by 38%.
• Developed real-time monitoring dashboards and experimentation frameworks enabling vehicle diagnostics, fleet optimization, model performance
• tracking, and engineering insights, reducing manual reporting efforts by 49% while supporting data-driven product development decisions.
• Collaborated with cross-functional engineering, automotive software, and operations teams to deliver connected vehicle solutions, integrating cloud
• infrastructure, embedded systems, and advanced analytics to enhance operational efficiency, system reliability, and customer experiences.