Looking for a job

Pavani Gulla

@pavanigulla

Message

I build production LLM, RAG, and real-time voice systems for regulated workflows.

United States

Message

What I'm looking for

I'm looking to build and deploy production AI systems, especially LLM, agentic, and real-time voice applications where rigorous evaluation, security, privacy, and scalable engineering matter.

At Harman International, I build large language model applications with retrieval-augmented generation pipelines, improving response accuracy by 32% through evidence-based iterative evaluation.

I create agentic workflows with DSPy and GEPA, optimize context engineering, and reduce latency for real-time voice systems. I also design secure agent orchestration and data stores for policy and coverage retrieval, including handoff to human agents.

I architect distributed systems with CI/CD automation, secure API integration, PII compliance, and release gates driven by error analysis, trajectory evaluation, golden sets, and LLM-as-judge loops.

Previously at Bosch Global Software Technologies, I developed and deployed machine learning models, secure claims-data pipelines, and evaluation tooling for production inference across 50,000 devices.

Experience

Work history, roles, and key accomplishments

Current

AI/ML Engineer

Current

Harman International

Sep 2024 - Present (2 years)

Developed large language model applications with retrieval augmented generation pipelines, improving response accuracy by 32% through evidence-based iterative evaluation. Built agentic workflows using DSPy and GEPA, and architected distributed systems with CI/CD automation.

Large Language Models RAG Agentic Workflows DSPy GEPA LLM As Judge Trajectory Evaluation Golden Set Evaluation Python CI CD Distributed Systems API Integration PII Compliance LiveKit SIP Real Time Voice Error Analysis Observability

ML Engineer

Bosch Global Software Technologies

Jul 2022 - Aug 2023 (1 year 1 month)

Developed machine learning and deep learning models with integrated testing and training frameworks, improving predictive accuracy on insurance claims data. Built secure data pipelines enforcing PII compliance and containerized ML models for scalable production inference.