Home

Neha B - Sr. AI Engineer

[email protected]

Location: Austin, Texas, USA

Relocation: YES OPEN FOR RELOCATION

Visa:

Resume file: Sr.AI Engineer Neha Bandari _1787775659066.docx Please check the file(s) for viruses. Files are checked manually and then made available for download.

NEHA BANDARI Sr. AI Engineer Email: [email protected] | Phone: +1 919-679-7812 | www.linkedin.com/in/neha3679 PROFESSIONAL SUMMARY: Senior AI Engineer with 11+ years of experience, progressing from software engineering and data engineering through machine learning to enterprise Generative AI and Agentic AI across regulated and industrial domains. Established software engineering and SDLC foundations at Medall Diagnostics by developing Python healthcare applications, REST APIs, SQL integrations, ETL pipelines, automated tests, documentation, and production-support processes. Advanced into AWS-based industrial analytics at AVEVA, combining Python, PySpark, scikit-learn, Amazon S3, Redshift, Flask services, and ETL processing for predictive-maintenance and equipment-health operational decision support. Delivered Azure data-engineering platforms for LA County using Azure Data Factory, Data Lake Storage, Synapse Analytics, PySpark, SQL, FastAPI, and governed batch and near-real-time reporting pipelines. Built BHG Financial credit-scoring, risk-modeling, and anomaly-detection solutions with XGBoost, TensorFlow, scikit-learn, Azure Machine Learning, MLflow, Databricks, and Snowflake on Azure for lending decisions and reporting. Architected Optum Generative AI, RAG, and Agentic AI workflows with Azure OpenAI, GPT-4o and GPT-4.1, LangChain, LangGraph, MCP, embeddings, vector search, tool calling, and evaluation for healthcare decision support. Engineered governed data workflows with Azure Data Factory, Azure Databricks, Apache Spark, PySpark, Pandas, Delta Lake, Snowflake, SQL, ETL/ELT, cleansing, validation, and quality controls for analytics and AI consumption. Constructed production software solutions with Python, FastAPI, Flask, Django, REST APIs, microservices, object-oriented design, exception handling, structured logging, secure coding, and reusable enterprise system integration patterns. Executed complete SDLC activities spanning requirements analysis, architecture, development, code reviews, unit and integration testing, deployment, Agile/Scrum delivery, documentation, user acceptance testing, monitoring, and production support. Automated version control and CI/CD workflows with Git, GitHub, GitHub Actions, Azure DevOps, Jenkins, AWS CodePipeline, Docker, Kubernetes, Terraform, and Infrastructure as Code across controlled release environments. Strengthened AI/ML quality through feature selection, hyperparameter tuning, prompt engineering, retrieval reranking, embeddings, vector search, model-drift monitoring, LLM evaluation, offline testing, and reproducible MLflow experiment benchmarking. Deployed observable cloud-native solutions on Azure and AWS with role-based access controls, HIPAA-aligned data handling, AI guardrails, Azure Monitor, Application Insights, Grafana, and CloudWatch supporting resilient production operations. TECHNICAL SKILLS: Generative AI & LLMs Azure OpenAI, GPT-4o, GPT-4.1, RAG, Agentic RAG, Agentic AI, LangChain, LangGraph, Model Context Protocol (MCP), multi-agent orchestration, tool calling, memory, prompt engineering, ReAct, embeddings, vector search, LLM evaluation, AI security guardrails Security & Governance HIPAA-aligned data handling, role-based access control, least-privilege access, audit logging, secure coding, AI security guardrails Machine Learning scikit-learn, XGBoost, TensorFlow, Spark MLlib, MLflow, Azure Machine Learning, feature engineering, feature selection, hyperparameter tuning, cross-validation, threshold calibration, classification, regression, anomaly detection, predictive analytics, model serving Programming & Practices Python, SQL, object-oriented programming, modular design, design patterns, requirements analysis, SDLC, exception handling, structured logging, secure coding, code reviews, Git Data Engineering & Modeling PySpark, Apache Spark, Pandas, Azure Data Factory, Azure Databricks, Snowflake on Azure, Delta Lake, ETL/ELT, data ingestion, batch and near-real-time processing, transformation, cleansing, data quality validation, star schema, dimensional modeling Cloud & Serverless Microsoft Azure, Azure Functions, Azure Service Bus, Azure Blob Storage, Azure Data Lake Storage, AWS EC2, Amazon S3, serverless architecture, event-driven architecture Data Storage & Retrieval Azure SQL Database, Azure Synapse Analytics, Snowflake, Delta Lake, Amazon Redshift, Azure AI Search, vector indexes, schema design, indexing, partitioning, query optimization Backend & APIs Python, FastAPI, Flask, Django, REST APIs, microservices, asynchronous processing, API integration, Swagger/OpenAPI, object-oriented design DevOps & Version Control Git, GitHub, GitHub Actions, Azure DevOps, Jenkins, AWS CodePipeline, Docker, Kubernetes (AKS), Terraform, Azure Resource Manager, Infrastructure as Code, CI/CD Monitoring, Testing & Delivery Azure Monitor, Application Insights, Grafana, AWS CloudWatch, model-drift monitoring, PyTest, unit testing, integration testing, A/B testing, user acceptance testing, production support, Agile/Scrum PROFESSIONAL EXPERIENCE: Role: Agentic AI Engineer Sep 2024 Present Client : UHG Optum, Eden Prairie, MN Responsibilities: Architected an enterprise Agentic AI platform for prior-authorization and claims triage with Azure OpenAI, LangChain, and LangGraph, supporting evidence-grounded clinical review across payer operations, appeals, and reviewer escalations. Mapped prior-authorization, claims-adjudication, member-eligibility, utilization-management, healthcare-interoperability, and clinical-data workflows, defining decision inputs, policy checkpoints, exception paths and human-review boundaries. Translated payer policies and clinical review needs into sequenced product capabilities, acceptance criteria, and release plans with Product Owners and subject-matter experts during Agile planning, refinement, demonstrations, and retrospectives. Designed event-driven microservices with Azure Service Bus, Azure Functions, and FastAPI to coordinate agent reasoning, tool execution, retries, asynchronous handoffs, and recoverable processing across long-running clinical review cases. Connected claims platforms, eligibility APIs, electronic-health-record interfaces, and Azure Blob Storage through validated ingestion services, enforcing schema checks, least-privilege access, and traceable handling of protected clinical documentation. Standardized claims documents, payer policies, and clinical narratives with Pandas, PySpark, and Azure Databricks, creating chunked, metadata-enriched content suitable for embeddings, filtering, retrieval, and downstream clinical reasoning. Organized structured case data in Azure SQL Database and Azure Synapse Analytics while maintaining Azure AI Search vector indexes with metadata filters supporting auditable retrieval across policy and case contexts. Evaluated approved GPT-4o deployments, prompts, embeddings, and retrieval configurations against grounding, latency, operational cost, and HIPAA-aligned handling requirements, then migrated the platform to GPT-4.1 once approved on Azure OpenAI, re-running grounding and regression evaluations before promoting the upgraded configuration to regulated production use. Orchestrated multi-step Agentic RAG flows through LangGraph state machines, enabling agents to retrieve policy evidence, query approved systems, preserve case state, and route uncertain determinations to human reviewers. Added MCP-based tool adapters during the later delivery phase, standardizing governed connections between agents and approved enterprise services after the protocol became available for production-oriented integration patterns. Engineered agent components with LangChain and object-oriented Python, standardizing prompts, tool contracts, state transitions, exception handling, structured logging, and reusable behaviors across deployable clinical workflow services. Developed secure FastAPI microservices exposing orchestration and retrieval capabilities through documented REST interfaces, request validation, authentication, correlation identifiers, and dependable integration contracts for upstream claims applications. Reduced unsupported responses through prompt refinement, embedding experiments, metadata filtering, and retrieval reranking, requiring policy citations and explicit evidence before automated recommendations reached clinical reviewers for determination. Enforced AI guardrails, role-based authorization, audit logging, and protected-data controls throughout agent workflows, restricting tool access and information exposure according to enterprise governance, privacy, and healthcare compliance requirements. Validated agent behavior with curated clinical scenarios, offline evaluation, adversarial tests, and reviewer feedback, scoring factual grounding, policy adherence, tool selection, and escalation behavior before promoting workflow changes. Containerized agent services with Docker and deployed them to Azure Kubernetes Service using autoscaling, rolling updates, readiness probes, and versioned configurations supporting dependable claims and utilization-management operations. Streamlined Git-based delivery through GitHub Actions and Azure DevOps pipelines covering code review, PyTest execution, security scanning, image publishing, environment promotion, and Terraform-based Azure infrastructure provisioning. Instrumented distributed workflows with Azure Monitor, Application Insights, and Grafana, tracing requests across agents and tools while reviewing latency, token usage, failures, retries, and retrieval-quality signals during operations. Authored Swagger specifications, architecture diagrams, evaluation records, and Confluence runbooks, supported structured user acceptance testing, and led knowledge-transfer sessions for engineering, clinical operations, and production-support teams. Environments: Azure OpenAI, GPT-4o, GPT-4.1, LangChain, LangGraph, RAG, MCP, Azure AI Search, Azure SQL Database, Azure Synapse Analytics, Azure Databricks, PySpark, Azure Service Bus, Azure Functions, FastAPI, Python, Docker, AKS, Terraform, Azure Monitor, Application Insights, Grafana, Git, GitHub, GitHub Actions, Azure DevOps, PyTest, Swagger, Agile/Scrum. Role: Applied AI/ML Developer Apr 2022 Aug 2024 Client : BHG Financial, Davie, FL Responsibilities: Built an Azure-based credit-scoring and anomaly-detection platform with XGBoost, TensorFlow, scikit-learn, and Azure Machine Learning, supporting risk modeling across loan processing, underwriting, portfolio monitoring, and servicing workflows. Clarified lending-policy, model-governance, and analyst-review requirements with Product Owners and risk stakeholders, converting business rules into feature definitions, acceptance criteria, validation plans, and controlled release milestones. Consolidated loan applications, borrower profiles, credit-bureau records, repayment history, and transaction data through Azure Data Factory pipelines with validation, reconciliation, retry handling, and traceable source-to-target mappings. Transformed raw lending data in Azure Databricks using PySpark and SQL, standardizing identities, resolving missing values, and deriving governed behavioral, affordability, repayment, and utilization attributes for modeling. Published curated lending datasets to Snowflake hosted on Azure, organizing role-governed schemas and reusable analytical tables for risk reporting, model-training extracts, portfolio analysis, and controlled downstream consumption. Developed reusable feature pipelines with Python and Spark MLlib, applying point-in-time logic and consistent transformations so training, validation, and production scoring consumed aligned borrower and loan attributes. Compared XGBoost, TensorFlow, and scikit-learn candidates using accuracy, interpretability, stability, and inference-latency criteria, prioritizing explainable models when credit decisions required defensible analyst review and governance approval. Applied classification, regression, and anomaly-detection methods to rank loan applications and suspicious activity by risk severity, model confidence, and analyst review priority across underwriting and servicing queues. Tuned hyperparameters, selected stable features, calibrated decision thresholds, and performed cross-validation across representative time-based samples, balancing credit-risk discrimination with manageable false-positive alert volumes for analyst investigation. Benchmarked candidate models against approved production baselines in MLflow, tracking experiments, parameters, artifacts, validation datasets, and comparison evidence required for reproducible governance review before controlled production promotion. Structured modular training, scoring, and evaluation packages with object-oriented Python, TensorFlow, scikit-learn, and Spark MLlib, enabling reusable pipeline components across multiple enterprise lending and portfolio-monitoring use cases. Designed FastAPI inference services and Azure Functions integrations that returned versioned risk scores and anomaly alerts to loan-origination and servicing applications with validated requests and graceful failure handling. Containerized training and inference components with Docker and deployed them on Azure Kubernetes Service, using autoscaling, versioned endpoints, health checks, and rollback procedures for dependable model serving. Automated governed MLOps workflows in Azure DevOps for feature preparation, training, validation, approval, deployment, and rollback, while Terraform and Resource Manager templates provisioned repeatable isolated Azure environments. Monitored feature distributions, prediction behavior, service health, and data-pipeline reliability with Azure Monitor, Application Insights, drift dashboards, and PyTest suites covering data transformations, APIs, and scoring logic. Documented model cards, endpoint contracts, Snowflake data mappings, validation evidence, and retraining procedures, then supported user acceptance testing and knowledge transfer with risk, analytics, engineering, and operations teams. Environments: Azure Machine Learning, MLflow, XGBoost, TensorFlow, scikit-learn, Spark MLlib, PySpark, Azure Databricks, Azure Data Factory, Azure Data Lake Storage, Snowflake on Azure, FastAPI, Azure Functions, Python, Docker, AKS, Terraform, Azure DevOps, Azure Monitor, Application Insights, PyTest, Agile/Scrum. Role: Sr. Python Software Engineer Apr 2019 Mar 2022 Client: LA County, Los Angeles, CA Responsibilities: Led Python backend and Azure data-engineering delivery for LA County reporting platforms, supporting governed ingestion, transformation, APIs, and analytics across public-sector programs, partner departments, and authorized downstream consumers. Refined reporting, audit, and data-delivery requirements with Product Owners and departmental stakeholders through Agile planning, backlog refinement, demonstrations, and retrospectives, producing sequenced backend and pipeline deliverables. Integrated case-management databases, departmental REST APIs, flat files, and relational sources through batch and near-real-time ingestion patterns, improving availability and freshness of program information for operational reporting. Established Azure Data Factory pipelines and Azure Data Lake Storage zones for raw, standardized, and curated datasets, applying parameterized workflows, schema validation, retry handling, access controls, and governed data movement. Standardized program data with Python, Pandas, PySpark, and SQL, resolving missing records, duplicate entities, schema drift, and inconsistent codes before information reached analytics, reporting, and audit consumers across county service programs. Modeled reusable star-schema fact and dimension tables in Azure SQL Database, aligning program, participant, service, and time definitions across departmental dashboards and recurring historical reporting workflows, enabling consistent public-service analysis. Expanded later project phases into Azure Synapse Analytics for scalable SQL and Spark processing, preserving governed source mappings while serving curated county datasets to reporting and analytical workloads. Developed documented REST services with Python, FastAPI, and Flask, connecting case-management applications, reporting tools, and downstream consumers through validated requests, consistent error handling, and maintainable integration contracts. Executed distributed ETL workloads with Apache Spark, PySpark, and SQL, processing recurring county datasets through reusable transformations, validation steps, and partition-aware writes for dependable scheduled delivery. Optimized Spark jobs and SQL queries through partitioning, indexing, caching, and execution-plan review, improving responsiveness and resource use across high-volume public-service reporting pipelines and analytical services. Verified completeness, reconciliation totals, referential integrity, and schema consistency before publishing curated data, documenting exceptions and preventing unreliable records from reaching dashboards, audits, and dependent application services. Containerized backend APIs and data services with Docker and deployed them on Azure Kubernetes Service, maintaining consistent configurations, health checks, and resilient hosting across development, testing, and production environments. Automated builds and deployments with Git, Jenkins, and Azure DevOps, while Terraform provisioned repeatable Azure resources for APIs, distributed data processing, monitoring, and isolated non-production environments. Instrumented pipelines and services with Azure Monitor and Application Insights, tracking run status, data-quality failures, API latency, and dependency errors to support proactive investigation and dependable county reporting operations. Maintained production pipelines, APIs, runbooks, and release documentation while supporting user acceptance testing, issue triage, and knowledge transfer for county analysts, application teams, and operational support groups across critical public-service programs. Environments: Microsoft Azure, Azure Data Factory, Azure Data Lake Storage, Azure SQL Database, Azure Synapse Analytics, Python, FastAPI, Flask, REST APIs, SQL, Pandas, PySpark, Apache Spark, Docker, AKS, Git, Jenkins, Azure DevOps, Terraform, Star Schema, Azure Monitor, Application Insights, Agile/Scrum. Role: Data Analytics Engineer Oct 2017 Mar 2019 Client: AVEVA, Lake Forest, CA Responsibilities: Built an AWS-based predictive-maintenance platform for AVEVA using Python, scikit-learn, and industrial sensor data, helping maintenance teams identify equipment degradation patterns and prioritize inspection before operational disruption. Translated reliability-engineering and plant-monitoring requirements with Product Owners and industrial specialists, defining sensor inputs, failure labels, analytical outputs, validation criteria, and incremental delivery priorities through Agile ceremonies. Centralized equipment readings, operating conditions, and maintenance history from industrial APIs and SQL sources into Amazon S3, applying repeatable ingestion schedules, source checks, and recoverable processing steps. Cleaned recorded high-frequency time-series data with Pandas and PySpark, handling missing intervals, outliers, inconsistent measurement units, and sensor-quality flags before statistical analysis and predictive-maintenance feature preparation. Designed Amazon Redshift tables and S3 storage layouts for retained sensor history, enabling efficient SQL analysis, trend reporting, and reproducible access to training and validation datasets for predictive-maintenance model development. Derived reusable equipment-health features with object-oriented Python, including rolling statistics, operating-range indicators, and change patterns that captured gradual degradation behavior across comparable assets and monitored industrial sites. Trained scikit-learn regression and classification models against historical maintenance events, comparing candidate approaches and selecting interpretable outputs that engineers could connect to equipment conditions and failure risk. Validated predictions through time-aware holdout datasets and offline backtesting, reviewing false alerts and missed events with maintenance specialists before integrating risk outputs into monitoring dashboards and alerting workflows. Developed Flask REST services that exposed approved predictions and equipment-health summaries to operational dashboards, using request validation, structured errors, and modular interfaces for dependable downstream integration. Accelerated recurring ETL workloads by tuning PySpark transformations and SQL queries, partitioning sensor data by asset and time, and reducing unnecessary scans across scheduled industrial analytics processing. Structured reusable ingestion, feature-engineering, scoring, and reporting modules in Python, clearly separating configuration from business logic and simplifying ongoing maintenance across development, testing, and production analytics environments. Containerized Flask analytics services with Docker and deployed them on AWS EC2, maintaining consistent runtime dependencies and repeatable configurations for development, testing, and production monitoring workloads. Automated source control, builds, tests, and releases through Git, Jenkins, and AWS CodePipeline, creating traceable deployment packages and controlled promotion paths for analytics services and scheduled jobs. Monitored service availability, pipeline executions, and infrastructure signals through AWS CloudWatch, configuring logs and alerts that supported proactive investigation across distributed industrial client environments and scheduled production operations. Documented data mappings, model assumptions, API contracts, deployment procedures, and operational runbooks, then supported release validation and production issue resolution with engineering, analytics, and maintenance stakeholders. Environments: Python, Flask, REST APIs, Pandas, PySpark, scikit-learn, SQL, Amazon S3, Amazon Redshift, AWS EC2, AWS CloudWatch, AWS CodePipeline, Docker, Git, Jenkins, statistical analysis, predictive analytics, ETL pipelines, Agile/Scrum. Role: Python Application Engineer Jul 2015 Aug 2017 Client: Medall Diagnostics, Chennai, India Responsibilities: Developed Python backend applications for diagnostic-lab operations and patient reporting at Medall Diagnostics, establishing foundational software-engineering experience across clinical data processing, laboratory workflows, and production support. Converted laboratory and reporting needs into sequenced application changes through Agile planning with Product Owners and operational teams, documenting requirements, acceptance criteria, dependencies, and release expectations. Structured modular backend services with Python, Flask, and Django for lab results, patient records, and diagnostic reports, separating business rules, data access, validation, and presentation responsibilities. Designed relational schemas and optimized SQL queries supporting diagnostic results and patient lookups, maintaining consistent keys, indexes, and retrieval patterns across high-volume daily laboratory reporting workflows. Built Python ETL pipelines that ingested lab results, patient records, and reference data from SQL sources and REST interfaces into validated reporting datasets and downstream operational applications supporting daily laboratory workflows. Validated diagnostic datasets with Pandas and schema-level checks, identifying missing values, duplicate records, invalid codes, and inconsistent formats before information reached laboratory reports and operational analytics. Delivered secure REST APIs with Flask and Django, enabling validated exchange of clinical and diagnostic information among internal laboratory systems, reporting tools, and approved external partner applications. Applied object-oriented Python and reusable modular design patterns to ingestion, validation, reporting, and integration components, improving maintainability and simplifying future enhancements across related diagnostic application modules. Implemented structured logging, exception handling, and input validation across backend services and ETL jobs, providing actionable diagnostics for production troubleshooting and dependable recovery from malformed or unavailable data. Executed PyTest unit and integration suites for business rules, API endpoints, SQL operations, and data pipelines, verifying application behavior before changes reached laboratory users and reporting workflows. Managed source changes with Git, participated in code reviews, and maintained release documentation, supporting traceable collaboration and controlled promotion of Python application updates across team environments. Deployed versioned application packages to Linux-hosted environments using standardized configuration and dependency-management procedures, maintaining runtime behavior across development, testing, and production diagnostic systems. Monitored application logs, scheduled ETL runs, and database connectivity, investigating failed records, delayed jobs, and service errors before they disrupted diagnostic reporting or daily laboratory operational workflows. Authored detailed API specifications, data mappings, support notes, and technical documentation, then assisted user acceptance testing and structured knowledge transfer with engineering, laboratory, and healthcare operations teams. Supported controlled production releases and incident resolution across patient-reporting services, diagnostic data pipelines, and backend integrations, coordinating timely fixes with application, database, laboratory, and downstream reporting stakeholders. Environments: Python, Flask, Django, Pandas, ETL pipelines, REST APIs, SQL, object-oriented programming, modular design, data validation, structured logging, PyTest, Linux, Agile, Git. EDUCATION: Bachelor of Technology - Computer Science and Engineering Keshav Memorial Institute of Technology (KMIT) Keywords: continuous integration continuous deployment artificial intelligence machine learning sthree California Florida Louisiana Minnesota