AI/ML Engineer - Multi-Agent Systems & LangGraph Architect

Location:

Austin, TX

Posted:

February 19, 2026

Contact this candidate

Resume:

Education

B. Arch(Bachelor in architecture ****-*004

UTKAL UNIVERSITY - India

Work Experience

Innovapath Inc - Dublin, CA May 2025- Present

AI/ML Engineer

Developed AI assistants using GenAI, LLMs, and multi-agent architectures to enable context-aware, personalized, and human-like conversational experiences. Engineered multi-agent systems enabling agents to plan, retrieve information, make decisions, validate actions, execute tools safely, track state, and collaborate on complex workflows. Built scalable agent workflows in LangGraph with branching logic, automated retries, and robust state management to ensure high-reliability across retrieval steps, tool interactions, and reasoning loops. Implemented stateful LangGraph agents with shared context, persistent state objects, and evolving memory structures to maintain coherent, long-running multi-turn conversations. Designed and deployed secure tool-use frameworks for AI agents using LangGraph and custom tool interfaces, ensuring safe and controlled external system interactions. Integrated Model Context Protocol (MCP) and secure tool interfaces to standardize agent-to-tool communication and enable safe access to enterprise systems. Architected a comprehensive memory system combining cache-backed short-term memory and database- backed long-term memory to deliver reliable and consistent conversational history management. Implemented full-stack observability for multi-agent systems using Prometheus and Grafana, enabling monitoring of prompt quality, performance metrics, and human feedback loops. Built reliability safeguards including fallback mechanisms, error-handling strategies, and safe-shutdown procedures to maintain agent stability during failures or model drift. Developed diagnostic pipelines to isolate issues across retrieval layers, agent graphs, LLM responses, vector databases, tool calls, and cloud environments (AWS EKS, Kubernetes), significantly improving debugging efficiency for complex agentic systems.

SUSHREE PRADHAN AI/ML ENGINEER

Austin, Texas +1-210-***-**** ************@*****.*** Linkedin USAA/HCL America- San Antonio, TX Jan 2024- April 2025 AI/ML Engineer

Designed advanced Agentic AI workflows using LangGraph and LangChain, enabling autonomous reasoning, dynamic tool coordination, and multi-step decision-making. Built stateful, modular agent architectures capable of handling evolving context, long-term goals, and adaptive strategy shifts for complex task execution. Integrated agents with external APIs, vector stores, and structured knowledge sources for real-time retrieval, contextual grounding, and dynamic reasoning. SUSHREE PRADHAN

Developed MCP-based connectors to support live data access and cross-system tool integration, boosting the agent’s situational awareness and task accuracy. Implemented RAG pipelines using Pinecone and Milvus to enhance relevance and knowledge augmentation for LLM responses.

Built evaluation systems using RAGAS, BERTScore, and structured metrics to measure retrieval quality, semantic alignment, and factual consistency.

Fine-tuned transformer models (e.g., BERT) using Hugging Face tools for classification and domain-specific NLP tasks.

Deployed scalable AI services on AWS (Bedrock, EKS) using FastAPI for real-time inference and optimized model serving.

Implemented end-to-end LLMOps pipelines with Docker and CI/CD to automate model versioning, deployment, and monitoring.

Optimized inference latency through efficient embedding handling, vector search tuning, and prompt strategies to improve real-time responsiveness.

Enhanced agent reasoning via multi-step orchestration, intelligent tool use, and adaptive behavior for complex query handling.

Built adaptive RAG architectures with feedback loops to continuously refine retrieval accuracy and contextual relevance.

Combined automated metrics with human evaluation to boost factual accuracy, coherence, and conversational quality.

Engineered robust prompt optimization and chain orchestration methods to improve reasoning depth and output reliability.

Ensured strong security, privacy, and compliance across all AI workflows and data interactions. Explored emerging generative AI techniques and integrated them to advance agent capabilities and system intelligence.

USAA/HCL America- San Antonio, TX Jan 2020 - Dec 2023 Machine Learning Engineer

Developed ML pipelines for clustering and segmentation of large-scale datasets, using AWS services to support efficient querying and seamless integration with ETL processes. Built scalable, automated workflows for data preprocessing, feature engineering, and model training with pandas, scikit-learn, and NumPy, ensuring consistent and high-quality transformations. Applied MLOps practices with MLflow, DVC, and Databricks to manage experiments, track datasets, and version models, improving collaboration and reproducibility. Deployed models using Docker, Kubernetes, and TorchServe on AWS SageMaker, delivering a robust and scalable production serving environment.

Designed CI/CD pipelines to automate testing, model deployment, and updates, reducing operational effort and improving release efficiency.

Created scalable data pipelines using AWS Lambda, S3, and related services to support real-time ingestion and smooth integration with machine learning workflows. Implemented monitoring and alerting solutions using CloudWatch and SageMaker Model Monitor to track model performance and ensure stability and uptime. Partnered with engineering and data teams to embed ML models into core systems, enhancing prediction accuracy, system performance, and overall business value. SKILLS

LLM, RAG, Vector DB, Langchain, Agentic AI

PyTorch, TensorFlow, Scikitlearn, MLOps, Hugging Face Python, JavaScript

PostgreSQL, MongoDB, MySQL

AWS, EKS, S3, SageMaker, Bedrock

HTML, CSS, JavaScript, React, Angular, Node.js, Redux Docker, Kubernetes, Jenkins, GitHub, FastAPI

Macy’s - San Francisco April 2018 - Dec 2019

Software Engineer

Designed, built, and maintained scalable software applications using contemporary programming languages and frameworks.

Created RESTful APIs and integrated external services to extend system capabilities and improve feature delivery.

Developed automated unit, integration, and functional tests to uphold code quality and reduce regression issues.

Worked closely with cross-functional teams to understand requirements and transform them into effective technical solutions.

Improved performance and reliability through continuous code reviews, optimization, and refactoring efforts.

Contributed to Agile/Scrum workflows, including sprint planning, task estimation, and daily stand-ups. Investigated and resolved complex software problems, delivering timely fixes and enhancements. Maintained clear documentation, technical specifications, and shared best practices to support team knowledge.

Supported deployment activities for both cloud and on-premises environments while adhering to security and compliance standards.

Wells Fargo, San Francisco CA Feb 2016 – Mar 2018