Agentic AI Orchestration & LLM Evaluation Expert

Pranav Gundaram

**************@*****.*** +1-443-***-**** LinkedIn GitHub

Summary

Agentic AI Expert with 5+ years of experience designing, orchestrating, and evaluating agentic AI workflows, LLM- based automation, and multi-agent systems using LangChain and LangGraph.

Proven track record delegating multi-step technical and software engineering tasks to AI systems, iteratively refining outputs, and evaluating AI-generated code for errors, hallucinations, and reasoning quality.

Skilled in prompt engineering, AI collaboration patterns, and building LLM evaluation frameworks (RAGAS, DeepEval) to benchmark output quality and reliability across systems.

Background spans software engineering, DevOps, and ML engineering – planning, implementing, debugging, and refactoring code in tool-integrated, agentic development environments.

Experienced integrating OpenAI, Anthropic (Claude), and Google Gemini APIs into production applications, with strong documentation practices for reproducible AI agent workflows.

Full-stack engineering foundation (FastAPI, Python, Docker/Kubernetes, CI/CD) supporting hands-on technical exe- cution, code review, and process documentation.

Technical Skills

Agentic AI & GenAI: Agentic AI Workflows, Multi-Agent Orchestration, LLMs, RAG, Prompt Engineering, Evaluation (RAGAS, DeepEval)

Frameworks & Libraries: LangChain, LangGraph, LlamaIndex, Hugging Face Transformers, OpenAI SDK, An- thropic (Claude), Gemini APIs

Languages: Python, Java, SQL, JavaScript, TypeScript Software Engineering: Code Review, Debugging, Refactoring, Technical Documentation, Reproducibility Backend & APIs: FastAPI, REST APIs, Microservices, Async Programming, API Orchestration, OAuth2/JWT

ML & Data Science: NLP, PyTorch, TensorFlow, scikit-learn, XGBoost, Spark (PySpark) DevOps & MLOps: Docker, Kubernetes, CI/CD (GitHub Actions, Jenkins), MLflow, Model Versioning, De- ployment Pipelines

Cloud Platforms: AWS (SageMaker, S3, Lambda), Azure OpenAI, Azure ML, GCP (BigQuery, Dataflow, Pub/Sub)

Monitoring & Observ-

ability:

Prometheus, Grafana, CloudWatch, Model Monitoring, Drift Detection Databases & Vector

Stores:

PostgreSQL, MySQL, Snowflake, Pinecone, FAISS

Frontend: React, Next.js, JavaScript, HTML/CSS

Professional Experience

Agentic AI Expert – Samsung Electronics USA Aug 2024 – Present

Built and orchestrated LangChain/LangGraph-based multi-agent workflows to automate fraud investigation and com- pliance analyst tasks, delegating multi-step technical work to AI agents and iteratively refining results.

Evaluated AI-generated code and agent outputs for errors, hallucinations, and reasoning quality, and documented best practices and process guidelines for reproducible agentic workflows.

Built RAG-based knowledge assistants enabling fast retrieval of regulatory and compliance documents using vector search, improving analyst response accuracy.

Designed low-latency inference pipelines on AWS SageMaker for real-time transaction scoring, integrated with agent- driven decisioning workflows.

Implemented end-to-end ML pipelines using MLflow and SageMaker Pipelines, applying LLM evaluation frameworks

(RAGAS/DeepEval) to assess output quality and production stability.

Developed NLP models (BERT/RoBERTa) for compliance analysis and risk-signal extraction, feeding structured out- puts into downstream AI agent tooling.

Developed secure FastAPI microservices with OAuth2/JWT authentication, enabling programmatic API access for AI coding agents and automation tools.

Environment: Python, FastAPI, LLMs, RAG, LangChain, Spark, AWS Applied AI Engineer – Vivian Contracting / Saint Agnes Jan 2024 – Jul 2024

Designed and built end-to-end ML pipelines for fraud detection, NLP, and forecasting use cases using Python, Spark, scikit-learn, TensorFlow, and XGBoost, improving model KPIs by 20%.

Developed scalable ETL pipelines using Spark and Apache Airflow to ingest, transform, and process 1M+ records/day.

Deployed containerized ML inference services and REST APIs on AWS SageMaker supporting 100K+ API requests/day.

Implemented MLflow for experiment tracking, model versioning, and reproducible deployments.

Built CI/CD pipelines reducing manual intervention by 40%.

Developed monitoring systems using Prometheus and Grafana for pipeline health and model drift.

Collaborated with cross-functional stakeholders to translate business requirements into ML solutions. Environment: Python, Spark, Airflow, AWS SageMaker, MLflow Software Engineer (ML) – Hitachi Vantara Aug 2020 – Jul 2023

Designed and implemented batch and real-time data pipelines using GCP services (Pub/Sub, Dataflow, BigQuery) to support trade analytics and risk monitoring.

Developed dimensional data models (Star/Snowflake Schema) and implemented Slowly Changing Dimensions (SCD Type I & II) to maintain historical data consistency.

Built machine learning models for fraud detection, customer segmentation, and forecasting using classification and clustering techniques.

Developed NLP-based classification systems to analyze compliance documents, KYC records, and customer communi- cations for risk insights.

Optimized SQL queries, indexing strategies, and data transformations to improve reporting performance and reduce processing latency.

Built dashboards and analytics solutions using Tableau and BI tools to provide actionable insights and support data governance initiatives.

Environment: Python, GCP (Pub/Sub, Dataflow, BigQuery), SQL, Spark, Tableau Projects

AI Agent Workflow Automation LangGraph, FastAPI

Built multi-agent orchestration workflows using LangGraph for enterprise automation use cases, delegating multi-step tasks to autonomous agents.

Developed tool-calling agents capable of interacting with APIs, databases, and external services dynamically, with iterative refinement of agent outputs.

Implemented memory and multi-step reasoning to maintain context across agent interactions, and evaluated outputs for accuracy and reliability.

Built an orchestration layer to manage communication and execution between agents at scale, with documented best practices for reproducibility.

Optimized pipelines using async execution, retry mechanisms, and error handling for production reliability. LLM-Based RAG System LangChain, Pinecone, OpenAI

Built an enterprise chatbot using LangChain and OpenAI APIs for intelligent knowledge retrieval from internal docu- ments.

Developed a full RAG pipeline including document ingestion, chunking, embedding generation, and vector indexing using Pinecone.

Implemented hybrid retrieval and re-ranking strategies to improve relevance of retrieved results and reduce hallucina- tions.

Designed prompt templates and grounding strategies for context-aware, reliable LLM responses.

Deployed solution using FastAPI with async processing and caching for low-latency concurrent access. Education

University of Maryland, Baltimore County (UMBC) Aug 2023 – May 2025 Master of Data Science, GPA: 4.0

Contact this candidate