Home

Mehdi Haghdad - AI/ML Architect

[email protected]

Location: Mission Viejo, California, USA

Relocation: No

Visa: USC

Resume file: Mehdi Haghdad Resume November 2025_1765031147529.doc Please check the file(s) for viruses. Files are checked manually and then made available for download.

Mehdi Haghdad Mission Viejo, CA 92692 [email protected] EDUCATION PhD, University of California UCLA/Davis 9/2000 - 6/2003 PhD in Electrical and Computer Engineering (Specializing in Advanced Artificial Intelligence, Neural Networks, Deep Learning, and Machine Learning for Smart Antenna Systems in Low Earth Orbit (LEO) Satellites). The Royal Institute Of Technology, Stockholm, Sweden M.S. Degree: Telecommunications Engineering 9/1998 - 6/1999 B.S. Degree: Computer Science 9/1995 - 6/1998 B.S. Degree: Electrical Engineering 9/1995 - 6/1998 COCIRTIFICATION AWS Certified Solutions Architect Professional May 2020. Microsoft Certified: Azure Solutions Architect Expert June 2021. Google Cloud Certified Professional Cloud Architect August 2022. SUMMARY I am a U.S. citizen with nearly 20 years of experience in advanced engineering 18 years (including my PhD) in Artificial Intelligence AI encompassing advanced machine learning (ML), deep learning (DL), convolutional neural networks (CNN, NLP, LLM, Generative AI etc. Throughout my career, I have worked with some of the most prestigious companies in Silicon Valley and beyond, including Lockheed Martin, Space Systems Loral, Microsoft (where I served four times as a Senior Solution Architect for various multibillion-dollar clients) , Amazon, Hewlett-Packard, Texas Instruments, Ericsson, T-Mobile, AT&T, Medtronic, Ernst & Young, Johnson & Johnson, Navy Federal Credit Union, Optum, UnitedHealthcare, Humana, Convey Health, Prudential Financial, Intuit, Acxiom, Citibank, Wells Fargo, Bank of America, TCF Bank, ABB Atom, Hilton, Wyndham Hotels, Adaptec, Broad Logic, Dell, Argonaut Technologies, Bausch & Lomb, DARPA, and the U.S. Department of Defense (DoD). I have led the development of numerous systems, guiding them from initial concept and architecture through to successful commercial deployment. OBJECTIVE Primarily seeking consulting and contract opportunities, but also open to permanent positions. EXPERIENCE Microsoft Client Sites West Coast to East Coast 9/2024-Present Senior AI/ML Architect | Hands-on Developer & Engineer This was a hands-on role where I architected, built, and deployed multi-cloud GenAI, Agentic AI, and RPA systems across Azure, AWS, and GCP for Microsoft enterprise customers. Designed and delivered AI applications such as intelligent email automation, AI copilots and virtual assistants (Microsoft Copilot, ChatGPT Enterprise, Gemini Advanced), RAG-based document intelligence, behavioral anomaly detection (SIEM), AI observability dashboards, and autonomous coding assistants (GitHub Copilot, CodeWhisperer). Architected distributed multi-agent systems using LangGraph, AutoGen, CrewAI 2.0, Semantic Kernel, DSPy, and Erlang/Elixir actor frameworks enabling task decomposition, Reflexion, Tree/Graph-of-Thought, and ReAct with interoperability layers powered by MCP (Model Context Protocol) and ACP (Agent Control Protocol) for agent coordination, context sharing, and secure multi-model orchestration. Implemented GraphRAG / GraphRAG++ architectures combining knowledge graphs (Neo4j, TigerGraph, ArangoDB, NebulaGraph) with hybrid RAG pipelines to enhance contextual retrieval and reasoning accuracy improving precision, scalability, and workflow autonomy. Multi-cloud GenAI platform: Built production ecosystems using Azure OpenAI, Amazon Bedrock, Google Vertex AI, Anthropic API, and Mistral API; integrated assistants (Copilot Studio, Azure AI Studio, Vertex Agent Builder, Dust.tt, Griptape, LlamaStack) and IaC (Terraform, Pulumi, CDK, Bicep) with DevOps (GitHub/GitLab CI/CD, CodeCatalyst). Agentic AI & autonomous systems: Developed multi-agent frameworks with LangGraph, AutoGen, CrewAI 2.0, Swarm, DSPy, AgentVerse, and LCEL; deployed on AKS / EKS / GKE / KNative with OPA/Kyverno policy controls, MCP-compliant adapters, and ACP-based control layers for dynamic agent collaboration and governance. GraphRAG & knowledge reasoning: Implemented GraphRAG pipelines using TransE/RotatE/ComplEx embeddings, hybrid retrieval (BM25 + vector fusion), and re-ranking (Cohere ReRank, Voyage ReRank, SPLADE); enhanced long-context compression (MemGPT, LongRAG). LLM API engineering & optimization: Built high-performance LLM APIs (FastAPI, vLLM 2.0, TensorRT-LLM, Triton, Ollama) with KV-cache and speculative decoding (Medusa, EAGLE, SpecInfer); quantized (AWQ, GPTQ, GGUF) on NVIDIA A100 / H100 / GH200 GPUs and orchestrated via Ray Serve, KServe, MLflow, Modal. Model alignment & evaluation: Fine-tuned LLMs with LoRA/QLoRA 2.0, PEFT, DPO/ORPO, RLHF/RLAIF, IPO/PPO; evaluated through LangSmith, LangFuse, PromptLayer 2.0, TruLens, Arize Phoenix 2.0 using guardrails (LlamaGuard 2, NeMo Guardrails, Presidio) aligned to NIST AI RMF and EU AI Act. Enterprise integration & automation: Connected Agentic and GraphRAG services to SAP S/4HANA/BTP, Dynamics 365, Salesforce, ServiceNow, and Workday via Kafka/Kinesis/PubSub/EventBridge and REST/gRPC APIs; automated flows with UiPath, Power Automate, Automation Anywhere, and Temporal.io. Document intelligence (IDP): Built OCR/IDP pipelines using Azure Document Intelligence, Amazon Textract, Google Document AI, LlamaParse, Docling, LayoutLMv3 for classification and workflow triggers. Models (frontier + OSS): GPT-4/4.1/4o/4.2, Claude 3.5 Sonnet/Opus, Gemini 1.5 Pro/Flash, Llama 3/3.1, Mistral Large/Mixtral, DBRX, Cohere Command-R+, Qwen 2.5, Phi-4, Falcon 180B, Yi-Large, SmolLM, Mamba-2; VLM/ASR models GPT-4V, Claude 3.5 Vision, LLaVA 1.6, BLIP-2, WhisperX. RAG stack & vector search: LangChain, LlamaIndex, Haystack, Semantic Kernel, AutoRAG; vector DBs Pinecone, Weaviate, Qdrant, Milvus, Chroma, FAISS, pgvector, Redis Vector, Vespa, OpenSearch Neural; embeddings text-embedding-004, BGE-M3, e5-mistral, voyage-large-3. Data & Analytics Platform: Azure: Microsoft Fabric, Unity Catalog, OneLake, ADLS, ADF, Databricks, Azure ML Studio, Power BI, Power Platform (Apps/Automate/Copilot). AWS: SageMaker, Bedrock, Redshift, EMR, Glue, Athena, S3, Aurora, Lambda, Fargate, ECS, Step Functions, Snowflake (Cortex, Snowpark, Streams & Tasks). GCP: Vertex AI (Agent Builder & Prompt Optimizer), BigQuery, Dataflow, Dataproc, Looker, TPUv5e. Search/Index: OpenSearch, Kendra, Elasticsearch ESRE. LLMOps / MLOps: Kubeflow, Ray, KServe, Airflow, Dagster, Flyte, BentoML, MLflow with LangSmith, LangFuse, Promptfoo, Evidently AI; safety via Azure AI Content Safety, Bedrock Guardrails, NeMo Guardrails, LlamaGuard 2; MCP/ACP observability extensions for cross-agent monitoring. Security & compliance: Zero-Trust architecture; IAM/AAD/GCP IAM; KMS/Key Vault/Cloud KMS; OPA/Kyverno/Gatekeeper; VPC/VNet isolation, PrivateLink, WAF; Confidential Computing (SEV-SNP, Nitro Enclaves, Confidential Space); model red-teaming and AI governance aligned with NIST AI RMF & ISO/IEC 23894. Data anomaly & drift detection: OpenTelemetry, Prometheus, Grafana, Flink, Spark Structured Streaming, Kafka Streams; detectors (Isolation Forest, RCF, Transformer Autoencoders); data quality via Great Expectations and Monte Carlo. Feature stores & governance: Feast, Tecton, Databricks/SageMaker Feature Store; governance with DataHub, Collibra, Atlan, Azure Purview; privacy via DP-SGD and synthetic datasets. Modernization & healthcare: Migrated SAP HANA/S4/BTP via SLT, CPI, BODS, Data Intelligence; streaming and batch with Kafka, Spark, Databricks; implemented FHIR/HL7 APIs and Microsoft Cloud for Healthcare meeting HIPAA, SOC 2, FedRAMP, ISO 27001. Edge & sovereign AI: Deployed models on Jetson Orin and ONNX Edge; enabled sovereign AI deployments (Azure Sovereign Clouds, AWS Sovereign Regions, PrivateGPT, LocalAI, Ollama Cloud). AI agents for DevOps & IT automation: Integrated AutoGPT Enterprise, DevOps Copilot, GitHub Copilot Workspace, and Jenkins AI Assist for CI/CD automation and incident triage. Representative Impact: Improved LLM grounding accuracy and retrieval fidelity by 40 60%; tripled automation throughput across ERP/CRM/ITSM via agentic workflows; cut inference latency ~50%; standardized AI observability and governance across multi-cloud environments. Amazon, Seattle, Washington 3/2024-9/2024 Senior AI/ML Architect | Hands-on Developer & Engineer Worked on the Alexa next generation, architected, created the necessary DevOps cloud infrastructure and developed (hands-on) the data pipeline and the AI/ML GenAI system to facilitate IoT hardware monitoring and data anomaly detection for the new generation of Amazon Alexa system. Designed and deployed advanced Generative AI (GenAI) and Retrieval-Augmented Generation (RAG) solutions on AWS using Amazon Bedrock, SageMaker, and OpenSearch with cutting-edge foundation models including GPT-4, Claude 3, Mistral/Mixtral, Cohere Command R+, Titan, and LLaMA 3. Architected scalable RAG pipelines leveraging vector databases (Pinecone, FAISS, OpenSearch), semantic search, and embedding generation with SageMaker JumpStart and LangChain. Integrated AWS services such as Lambda, API Gateway, DynamoDB, and Step Functions to orchestrate GenAI workflows with secure, real-time document retrieval and context-aware LLM interactions. Built multi-agent reasoning systems using CrewAI and LangGraph, enabling complex task automation and dynamic memory management. Implemented prompt engineering, safety guardrails, and model evaluation with PromptFlow, Traceloop, and Ragas. Delivered enterprise-grade GenAI platforms with secure access control (IAM, KMS), logging (CloudWatch), and CI/CD automation using CodePipeline and Terraform. Architected, hands-on developed and delivered enterprise-grade data solutions on AWS leveraging modern services including Redshift Serverless, EMR on EKS, Aurora (MySQL/PostgreSQL), S3, Glue Studio, Glue Data Catalog, Athena, and DMS for robust ingestion, transformation, and analytics. Built end-to-end scalable data pipelines using Spark on EMR, Kafka (MSK and Confluent), and AWS Data Pipeline to process structured and unstructured data in real-time and batch modes. Implemented Medallion architecture (Bronze, Silver, Gold layers) on S3 and Lake Formation for scalable, modular, and secure data lake design. Integrated data governance platforms like Collibra, Alation, and Atlan with AWS Glue and Lake Formation for lineage, metadata management, and policy enforcement. Enabled advanced analytics and BI through Amazon QuickSight and Redshift ML, supporting interactive dashboards and machine learning inference on data at scale. Ensured high availability, cost optimization, and automation through Infrastructure as Code (Terraform, CloudFormation) and CI/CD pipelines for seamless deployment and lifecycle management. AT&T, Dallas Texas 2/2022-3/2024 Senior AI/ML Architect | Hands-on Developer & Engineer At AT&T, served as a Senior AI/ML Architectbut as a hands-on senior developer writing code, creating cloud infrastructure DevOps, participating in testing and deployment, I led, designed and deployed Generative AI and RAG-powered solutions on AWS to support 5G anomaly detection, predictive maintenance, and intelligent customer service automation. Architected and deployed Retrieval-Augmented Generation (RAG) systems using FAISS, Dense Passage Retrieval (DPR), Siamese BERT, OpenSearch, and Amazon Kendra, enabling context-aware Q&A and dynamic support insights by fusing dense vector retrieval with transformer-based response generation. Integrated state-of-the-art LLMs including GPT-3, GPT-Neo, BERT, T5, RoBERTa, DistilBERT, XLNet, ELECTRA, ALBERT, CTRL, and GPT-J using Hugging Face Transformers, SageMaker JumpStart, and OpenAI API, to power summarization, log understanding, chatbot response synthesis, and dynamic document generation. Engineered generative NLP pipelines for multilingual support, sentiment analysis, and real-time diagnostics using Amazon Comprehend, Translate, Lex, and custom-tuned LLMs deployed via SageMaker Endpoints. Employed Prompt Engineering techniques to fine-tune generation outputs for different customer service scenarios and technician assistance. Developed full-stack MLOps workflows with SageMaker Pipelines, Model Registry, and Model Monitor to automate model training, evaluation, deployment, and drift detection. Managed lifecycle with MLflow, and ensured model explainability using SHAP and LIME to meet enterprise AI governance requirements. Ingested real-time 5G telemetry and user interaction data using AWS IoT Core, Amazon Kinesis, and Kafka (MSK), triggering downstream GenAI model activations via Step Functions and Lambda, with outputs visualized in Amazon QuickSight and Plotly Dash. Enabled scalable and reproducible infrastructure using AWS CDK and CloudFormation, and built CI/CD automation via CodePipeline, CodeBuild, and GitHub Actions. Applied security best practices using IAM, KMS, and Lake Formation, achieving compliance with SOC 2 and ISO 27001. Delivered transformer-based anomaly detection using Prophet, DeepAR, and XGBoost, with generative narrative reporting of anomalies for technical staff. These systems reduced mean time to resolution by over 40% and automated root-cause diagnostics across AT&T s 5G infrastructure. Led the development of an LLM-based digital assistant framework for Tier-1 support, combining Lex, RAG architectures, GPT-Neo, and SageMaker, resulting in a 35% increase in first-contact resolution and significant reductions in support ticket backlog. Microsoft, New York, Newport Beach California 3/2019-2/2022 Senior AI/ML Architect | Hands-on Developer & Engineer As a Solution Architect, high-level lead, and expert. I collaborated with numerous prominent Microsoft clients, primarily Fortune 500 companies, to enhance their existing architectures or design entirely new ones. My role encompassed guiding clients through cutting-edge technologies and implementing effective solutions. Below are some of the key areas I have worked in: As a Senior AI/ML Architect | Hands-on Developer & Engineer, I designed, architected, and led enterprise-scale AI initiatives across on-prem, AWS, Azure, and GCP environments, delivering real-time credit auditing, CRM automation, intelligent email classification, AI auto-reply, and anomaly detection systems. Leveraging Convolutional Neural Networks (CNNs), transformer-based models, and multi-agent systems, I deployed scalable Generative AI and Retrieval-Augmented Generation (RAG) architectures powered by GPT-2, GPT-3, T5, BERT, RoBERTa, DistilBERT, XLNet, ELECTRA, and ALBERT. I implemented RAG pipelines using FAISS, Haystack, OpenSearch, Hugging Face Transformers, and Amazon Kendra, enabling semantic search, document summarization, and contextual Q&A. I integrated embeddings from OpenAI, Sentence-BERT, and Universal Sentence Encoder to support deep similarity search and hybrid retrieval. GenAI deployments were unified using SageMaker, Azure ML, and Vertex AI, orchestrated via LangChain, FastAPI, and Airflow, and scaled with Kubernetes, Docker, Terraform, and Kubeflow for CI/CD and model serving. Optimized transformer inference using NVIDIA V100/A100 GPUs, TensorRT, ONNX Runtime, and DeepSpeed, while applying Low-Rank Adaptation (LoRA) and knowledge distillation for model compression. Enabled multi-agent collaboration through early frameworks like Haystack pipelines and custom agent routing, and integrated Explainable AI (XAI) via SHAP and LIME. Applied RLHF, Chain-of-Thought prompting, and ethical AI design patterns to ensure transparency, safety, and alignment with responsible AI principles. In AWS, Lead, architected and developed a comprehensive AWS infrastructure with: Redshift (also tested POC in Snowflake Data Warehouse and Snowpipe), RDS, EMR, MSK (Kafka Confluent), IoT (Internet of things), SCADA, PLC, S3 AWS Compute E2C, Glue, AWS Connect, DMS, Athena, EC2, RDS Aurora MySQL PostgreSQL, Couchbase, MemSQL, Elasticsearch, Lambda projects with full Big Data, Amazon Elastic MapReduce (EMR), CI/CD (CICD, CI CD,CI-CD) pipeline, CloudFormation CFN, Spark, Kafka Confluent, Oozie, Sqoop, NiFi, Pig, Hive, Hbase, MSK, AWS CLI, Amazon EMR File System (EMRFS). Comprehensive CDK and CloudFormation experience with writing script in TypeScript, JavaScript, Python, Java, and C#. Collaborative notebooks Zeppelin, Jupyter. AWS networking used VPC, Private Subnet, Public Subnet, Internet Gateway, Routing Security groups etc. Visualization and analysis used QuickSight and CloudSearch. Also created a Kubernetes cluster using Elastic Container Service for Kubernetes (EKS), App Mesh, EC2 Container Service (ECS), FarGate. For AI used deep learning frameworks like Apache MXNet Machine Learning and Deep Learning development and deployment, VMware, AWS Developer Tools, AWS Management Tools, Amazon Machine Learning, SageMaker, Alexa Skills Kit, AWS DeepLens, Amazon Deep Learning AMIs, Amazon TensorFlow on AWS, Amazon CodeWhisperer AI and other components. For security used Identity and Access Management (IAM), AWS Organizations, Multi-Factor Authentication, SAS, migration of On-Prem Windows SAS to AWS SAS Studio etc. Also lead, architected and helped developing multiple CDK and CloudFormation projects with writing script in TypeScript, JavaScript, Python and Java. In Azure, designed and built an end-to-end enterprise pipeline using Azure Data Factory Gen2 to orchestrate ingestion from SQL, flat files, Apptio Cloudability, data warehouses, and external sources. Architected a 4-zone ADLS Gen2 data lake (Raw, Staging, Curated, Consumption) aligned with Microsoft s best practices. Streamed critical data via Kafka Confluent and less critical via Azure Event Hub into the Raw Zone. Integrated Delta Lake, supporting table versioning, schema evolution, and ACID transactions. Configured Azure Databricks clusters with Unity Catalog for workspace-level governance, access control, lineage, and auditing. Managed model lifecycles using MLflow, Unity Catalog, and REST API endpoints for deployment and tracking. Built real-time and web apps with .NET, Angular, React.js, Spring Boot, and Node.js, deployed inside and outside AKS with APIM and Service Mesh. Designed dashboards in Power BI, integrating with Synapse Analytics, Microsoft Fabric Lakehouses, OneLake, and Azure Purview for metadata and cataloging. Enhanced cloud security with Microsoft Sentinel, Azure AD, AADDS, Key Vault, Service Principals, Private Links, and Azure VNet. Developed LLM-powered solutions using Azure OpenAI, ChatGPT API, Prompt Engineering, and Microsoft Bot Framework for sentiment analysis and automation. Implemented CI/CD pipelines using Azure DevOps, GitHub Actions, Azure Boards, and Azure Application Insights. Automated infra-as-code with ARM Templates, Terraform, and Visual Studio. Designed data warehousing in Azure Synapse and Snowflake, with optimized pipelines and DWH performance tuning. Enabled HIPAA-compliant FHIR/HL7 integration using FHIR APIs, REST services, and JSON for EHR interoperability. Delivered full-stack observability and analytics using Highcharts, Encore Analytics Empower, and custom telemetry dashboards. In GCP, worked on different component of the GCP among others Cloud Dataproc, BigQuery, Airflow Cloud Dataflow, Cloud Data Fusion, Cloud Dataprep, Data Catalog, Google Kubernetes Engine (GKE), Kubernetes CLusters, Dataproc CLusters, Container Registry, CI/CD (CICD, CI CD,CI-CD) pipelines, Deep Learning Containers, Cloud Bigtable, Cloud SQL, Firebase Realtime Database. For artificial intelligence used AI building blocks, Text-to-Speech, Speech-to-Text, AutoML, Vision AI, Vertex AI, Cloud Natural Language, Video AI, AI Platform, AI Hub and AI Platform Deep Learning VM Image etc. Ericsson, Santa Clara, California 8/2018-3/2019 Senior AI/ML Architect | Senior Solution Enterprise Architect | Hands-on Developer & Engineer At Ericsson s Global AI Accelerator (GAIA) in Santa Clara, I served as a Senior AI/ML Architect and Lead Data Engineer, leading the development of a $75M/year initiative to revolutionize 5G telecommunications using Machine Learning (ML), Deep Learning (DL), and Convolutional Neural Networks (CNNs). I architected and deployed a 2,000-node, 16,000-core Kubernetes cluster handling 60TB/day of real-time data for advanced wireless analytics. This infrastructure supported real-time anomaly detection, system performance prediction, and cybersecurity threat detection using deep learning models built with TensorFlow, Keras, MXNet, and Scikit-learn. The project included high-throughput pipelines converting signal telemetry into image-like formats processed by CNNs similar to early YOLO and AlexNet architectures for enhanced accuracy. My role involved building enterprise-scale AI infrastructure across AWS, Azure, and OpenStack. I implemented petabyte-scale data lakes with HDFS, Hive, and Spark, integrated with Kafka Confluent, NiFi, and ELK stack on Kubernetes. I used Terraform and Vagrant to deploy reproducible environments and built distributed ML systems using Scala (SBT), Python, and Zeppelin/Jupyter notebooks. Model development leveraged classic algorithms like Random Forest, Multilayer Perceptron (MLPC), Gradient Boosting (XGBoost, GBM), and k-Means, with GPU acceleration where applicable. CNN pipelines were enhanced through Acumos AI (early stages), Caffe2, Deeplearning4j, and Theano, with experimentation in cascading neural networks for modular AI design. Tools like R, PySpark, and Java were used for cross-platform integration and experimentation. Within AWS, I developed Big Data pipelines using S3, EC2, EMR, Redshift, RDS, Kinesis, Glue, and Lambda for ingestion, transformation, and analytics. AI workloads were trained and deployed using Amazon SageMaker (introduced late 2017), enabling distributed model training, endpoint inference, and hyperparameter tuning. I implemented CI/CD workflows using Jenkins and GitLab CI with automated deployment into EMR and SageMaker. Visualization was handled via Amazon QuickSight and custom dashboards using Kibana and Zeppelin notebooks. I leveraged AWS ML services including Amazon Machine Learning and built predictive systems that interfaced with AWS SDKs, EC2 Auto Scaling, IAM, and CloudWatch for monitoring. This AWS-native stack allowed scalable development, deployment, and monitoring of AI/ML solutions in production-grade environments. Developed GPU-accelerated AI pipelines using NVIDIA Tesla V100 and P100 GPUs with CUDA, cuDNN, and NCCL for high-performance model training. Deep learning models were implemented using TensorFlow, Keras, MXNet, and Caffe2, optimized for parallel training on multi-GPU clusters. I applied CNNs, RNNs, and LSTMs for time series prediction and pattern recognition in wireless data. Leveraged NVIDIA DIGITS for rapid prototyping and model benchmarking. Used Docker and NVIDIA-Docker for GPU-based containerization and deployment. Integrated OpenCV and CUDA-accelerated libraries for image-based anomaly detection in streaming 5G signal data. Microsoft, New York, California 3/2018-8/2018 Senior AI/ML Architect | Senior Solution Enterprise Architect | Hands-on Developer & Engineer As a Senior Solution Architect Consultant and Senior Developer, I led the architecture, design, and hands-on implementation of a large-scale Azure Cloud project in collaboration with Microsoft, Pragmatic Works, and Selective in New York. I provided deep expertise in Azure Big Data, ETL, advanced Machine Learning (ML), and Deep Learning (DL), working from initial design through production release. I architected end-to-end data pipelines covering ingestion, transformation, and consumption using tools like Azure Data Factory, Azure Synapse Analytics, and Azure Databricks. Leveraging Spark, HDInsight, and Data Lake, I integrated advanced AI neural network models via Azure Machine Learning Studio and Microsoft R Server. I ensured secure, scalable infrastructure through Azure Active Directory, Service Principals, Key Vaults, and Azure Kubernetes Service (AKS). This hands-on project covered full-stack Azure cloud architecture including Power BI, Azure Search, Cosmos DB, Azure SQL Data Warehouse, Snowflake, and Elasticsearch on Azure. I developed and deployed scalable infrastructure using ARM Templates, Azure Runbooks, Visual Studio 2017, and integrated CI/CD pipelines with Azure DevOps. I built HDInsight clusters, including domain-joined configurations, and worked with Zeppelin and Jupyter Notebooks on Azure. I handled real-time streaming via Kafka, Confluent, and Event Hubs, and implemented polybase ingestion, text analytics, key phrase extraction, and Cortana Intelligence Suite. My experience includes Azure Log Analytics, Network Interfaces, Azure Integration Runtime, Azure Storage Blobs, Virtual Networks (VNet), and Azure Stack HCI, delivering both IaaS and PaaS solutions at scale. Optum Health United Health Care, Santa Ana (partially in Minneapolis), California 8/2016-3/2018 Senior AI/ML Architect | Senior Solution Enterprise Architect | Hands-on Developer & Engineer As a Lead Senior Solution Architect Consultant, I led and implemented cutting-edge AI/ML, Big Data, and cloud solutions for Optum and other major clients, including Microsoft and IBM. I architected scalable distributed systems with 12,000 CPUs for genomic analysis, reducing compute time from 46 years to under 23 minutes with 99%+ cancer prediction accuracy, using GATK4, Berkeley AMPLab ADAM Genomics, Spark, HDFS, Kafka, TensorFlow, Keras, MLlib, XGBoost, PySpark, R, Scala, and Julia. I developed advanced DL models (CNNs, MLPC, RF, GBM) for cancer prediction and high-speed query processing across 4 billion records. My solutions spanned Azure, AWS, and IBM Cloud, combining Kubernetes, HDInsight, Databricks, and Azure ML Studio for full-cycle ML workflows. Built hybrid data ingestion pipelines using Kafka, Hive, HBase, and MarkLogic, and led real-time search systems on Elasticsearch (25-node cluster), Kibana, and Solr, integrated via Logstash, Rsync, and REST APIs. My cloud deployments included Azure Synapse, Azure Data Lake, Azure Key Vaults, ARM templates, Azure Runbooks, AKS, Cosmos DB, Power BI, Azure Search, and CI/CD pipelines via Azure DevOps. On AWS, I architected solutions using EMR, Glue, S3, Athena, Lambda, MXNet, and Amazon SageMaker. I deployed ELK stacks and Kafka ecosystems on IBM Cloud Private (ICP) using Helm, Docker, and Kubernetes, and adhered to strict HIPAA and FHIR HL7 standards for healthcare compliance. My preferred languages include Scala, Python, Java, and Shell, with deep experience using IntelliJ, SBT, Jupyter, and Eclipse in high-performance, enterprise-grade environments. Other Projects Between 4/2006-9/2016 OneStop, El Segundo, California 2/2016-8/2016 Senior lead, Senior Solution Architect Consultant, senior developer, Big Data, Data Warehousing, BI, SOLR, Lucene, Elasticsearch, Mahout, Weka Machine Learning Lead. NovaWurks/ DOD DARPA, Los Alamitos, California 11/2014-2/2016 Senior Big Data, DW and BI Lead Solution Architect Consultant, Java Android consultant Paramit, Morgan Hill, California 7/2013-11/2014 Senior Big Data, DW and BI Lead Solution Architect, .NET Architect Consultant. Microsoft, Redmond WA 2/2013-7/2013 Senior Big Data, DW and BI Lead Solution Architect, .NET Architect Consultant. Dell, Austin, TX 6/2012-2/2013 Senior Big Data, DW and BI Lead Solution Architect, .NET Architect Consultant. BEW / General Electric / 3 Gorges China, San Ramon, California 3/2011-6/2012 Team Leader, .NET Architect, Hands on Developer Consultant. Texas Instruments, Dallas Texas 9/2010-3/2011 Team Leader, .NET Architect, Hands on Developer Consultant. Multibeam Corporation / Tokyo Electron (TEL), Santa Clara, California 3/2010-9/2010 Team Leader, Architect, Hands on Developer Consultant. Department of Defense Contract (DOD), Washington DC 9/2009-3/2010 Team Leader, Architect, Hands on Developer Consultant. Hewlett Packard (HP), Cupertino, California 3/2009-9/2009 Team Leader, Architect, Hands on Developer Consultant. Broad Logic, Milpitas, California 8/2009-3/2009 Team Leader, Architect, Hands on Developer Consultant. Hewlett Packard (HP), Mayfield, California 2/2009-8/2009 Team Leader, Architect, Hands on Developer Consultant. Hewlett Packard (HP), Cupertino, California 10/2008-2/2009 Team Leader, Architect, Hands on Developer Consultant. Space Systems Loral (CyberStar), Mountain View, California 10/2007-10/2008 Team Leader, Architect, Hands on Developer Consultant. Lockheed Martin, Milpitas, California 4/2007-10/2007 Team Leader, Architect, Hands on Developer Consultant. Ericsson (Ellemtel), Stockholm, Sweden 8/2006-4/2007 Team Leader, Architect, Hands on Developer. ABB Atom AB, Vasteras, Sweden 2/2006-8/2006 Team Leader, Architect, Hands on Developer Consultant. SKILLS Summary of Technical Expertise With 20+ years of experience and a PhD from University of California UCLA/Davis, I m a Senior Solution/Enterprise/Technical Architect, Engineering Leader, and hands-on Developer across AI/ML, GenAI, Cloud, Big Data, Kubernetes, DevOps, Data Warehousing, and Enterprise Systems (AWS, Azure, GCP, on-prem, hybrid). ________________________________________ GenAI, LLMs & Agentic AI Models (frontier + OSS): GPT-4/4.1/4o/4o-mini, Claude 3/3.5 (Opus/Sonnet/Haiku), Gemini 1.5 Pro/Flash, Llama 2/3/3.1, Mistral Large/Mixtral/Codestral, DBRX, Cohere Command-R/R+, Amazon Titan, Qwen 2/2.5, Phi-3/4, Falcon, PaLM 2, XGen, Zephyr, OpenChat, DeepSeek-VL/DeepSeek-Coder, WizardLM, Mosaic MPT, Vicuna, Yi, Open WebUI. VLM/ASR: LLaVA, BLIP-2, CLIP, Whisper/WhisperX, OpenAI Voice Engine. Agent frameworks & reasoning: LangGraph v2 (state, memory/checkpoints), CrewAI v2, AutoGen, Semantic Kernel, DSPy, OpenDevin, SuperAGI, Toolformer, AgentOps, MemGPT; patterns ReAct, Reflexion, Tree/Graph-of-Thought, ToD tool use; JSON Schema/OpenAPI tools; Databricks Mosaic AI (Vector Search, Agent Framework); NVIDIA NIM; NeMo Guardrails. Assistants & studios: Microsoft Copilot Studio, Azure AI Studio, Vertex AI Agent Builder, Open WebUI; multi-cloud model access via Azure OpenAI, Amazon Bedrock, Vertex AI Model Garden/Google Model Garden. RAG & knowledge reasoning: LangChain, Haystack, LlamaIndex, LCEL; GraphRAG with knowledge graphs (Neo4j, TigerGraph), KG embeddings (TransE/RotatE/ComplEx); hybrid search (BM25+vector), re-ranking (Cross-Encoders, SPLADE, Cohere/Voyage/Jina ReRank, ColBERT-v2), query expansion (HyDE), session memory, long-context; tool-augmented verification and grounded generation. Vector DBs & embeddings: Pinecone, Weaviate, Qdrant, Milvus, Chroma, FAISS, pgvector, Redis Vector, Vespa, OpenSearch/Elasticsearch vector, Meilisearch, Typesense, LanceDB, Vectara; embeddings text-embedding-3-large/small, text-embedding-004, BGE-m3, e5-mistral/e5-large-v2, nomic-embed-text, voyage-large-2/3, multilingual-e5/bge; tokenization Hugging Face Transformers, OpenAI/Cohere embeddings, FastEmbed, InstructorXL. ________________________________________ LLM API Engineering, Training, Efficiency & Accuracy Serving & APIs: FastAPI, vLLM (paged attention), Text Generation Inference (TGI), TensorRT-LLM, NVIDIA Triton, SGLang, Ollama, TorchServe, KServe, Ray/Ray Serve, MLflow; continuous batching, KV-cache, prompt caching (Redis/pgvector), speculative decoding (Medusa, ReDrafter, SpecInfer). Perf/efficiency: FlashAttention-2/3, DeepSpeed, FSDP, ZeRO/ZeRO-Infinity, Megatron-LM (tensor/pipeline parallel), CUDA Graphs, TorchInductor/TorchCompile, ONNX Runtime, OpenVINO, CTranslate2, FasterTransformer; quantization AWQ/AutoAWQ, GPTQ, GGUF, EXL2; FP16/BF16/FP8/NF4. Fine-tuning & accuracy: SFT, LoRA/QLoRA/PEFT, DPO/ORPO, RLHF/RLAIF, IPO; synthetic data (SDV/YData); self-consistency/CoT, calibration & uncertainty, LM-Eval-Harness; XAI (SHAP, LIME, Captum). Hardware & infra: NVIDIA A100/H100, DGX, NGC/NIM, AWS Trainium/Inferentia, Habana Gaudi, Azure NDv5, Google TPU v4/v5e. ________________________________________ MLOps / LLMOps, Observability & Guardrails Pipelines & registries: Kubeflow, SageMaker Pipelines, Azure ML Studio, Vertex AI Model Registry, Azure Databricks (Unity Catalog), PromptFlow, MLflow, Hugging Face TGI, KServe, Seldon Core, BentoML, Airflow, Dagster, Flyte, Backstage; data/model versioning LakeFS, DVC. CI/CD & artifacts: Jenkins, GitHub Actions, GitLab CI/CD, Bitbucket, ArgoCD, FluxCD; Docker, Helm, Bitnami, JFrog Artifactory. Eval/telemetry/guardrails: LangSmith, Ragas, TruLens, Promptfoo, DeepEval, Arize Phoenix, Langfuse, Helicone, Weights & Biases, Evidently, Traceloop, OpenLLMetry; safety/PII: Azure AI Content Safety, Bedrock Guardrails, Google Safety Filters, NeMo Guardrails, Presidio. ________________________________________ Security, Responsible AI & Compliance (incl. Model Security) Frameworks & risk: OWASP Top-10 for LLMs, NIST AI RMF, ISO/IEC 23894, Responsible AI (RLHF, CoT, safety classifiers, Trust Layers), Model Cards. Controls & runtime: Prompt-injection/jailbreak detection, output filtering, data exfiltration controls, model red-teaming/adversarial testing, DLP/PII (Presidio, Cloud DLP), supply-chain (model SBOM, checksums, provenance). Platform security: IAM/AAD/GCP IAM, RBAC/ABAC, KMS/Key Vault/Cloud KMS, OPA/Gatekeeper/Kyverno, Zero-Trust, TLS/mTLS, OIDC/JWT/OAuth2, Binary Authorization, service mesh (Istio, Linkerd), Private Link/Endpoints, SCIM. Confidential computing: Azure SEV-SNP, AWS Nitro Enclaves, GCP Confidential Space; envelope encryption & key rotation. Compliance & governance: SOC 2, ISO 27001, HIPAA, GDPR, NIST 800-53/FedRAMP; Azure Purview, Amazon Macie, Google DLP, Secure Score, TrustArc, Immuta, SAP GRC. ________________________________________ Real-Time AI, Fraud & Anomaly/Drift Detection Streaming & stores: Kafka (Confluent/MSK), Kinesis, Pub/Sub, Azure Event Hub, Spark Structured Streaming, Apache Flink, Kafka Streams; Pinot, Druid, ClickHouse. Detectors & libs: Isolation Forest, One-Class SVM, LOF, Robust Random Cut Forest, ARIMA/Prophet, Matrix Profile, S-H-ESD, LSTM/Transformer autoencoders, DAGMM; PyOD, Merlion, River. Quality, lineage & contracts: Great Expectations, Deequ, Monte Carlo, Soda, OpenLineage/Marquez. Observability: Prometheus, Grafana, OpenTelemetry, CloudWatch, Azure Monitor, GCP Operations Suite, SageMaker Model Monitor; SIEM: Microsoft Sentinel, GuardDuty, Cloud IDS. ________________________________________ Multimodal: Speech, Vision & Document AI Vision: YOLOv8, ViT, SAM/SAM2, Mask R-CNN, GroundingDINO, OWL-ViT, Florence-2, CLIP, BLIP. Generative image/video: Stable Diffusion 3, DALL-E 3, Midjourney, Ideogram. Speech & realtime: Whisper, Riva, Microsoft Speech Studio, Azure Cognitive Services, Amazon Transcribe, Google Speech-to-Text; Realtime APIs (OpenAI/Vertex/Azure). Document AI: LayoutLMv3/Donut/TrOCR, Azure Document Intelligence, Amazon Textract, Google Document AI, Unstructured, LlamaParse, Docling. ________________________________________ Data Engineering, Big Data & Streaming Core stack: Hadoop, Spark, Storm, Kafka (Confluent/MSK), Flume, Sqoop, HBase, NiFi, Hive, Pig, Impala, MapReduce, Delta Lake, Databricks, Lakehouse, OneLake, PolyBase, IoT Core, SCADA, PLCs. Tables, CDC & EL(T): Apache Iceberg, Apache Hudi, Delta Live Tables; Airbyte, Fivetran, Debezium, Kafka Connect. Query & engines: Trino/Presto, ClickHouse, Pinot, Druid, DuckDB. Catalog, governance & lineage: Collibra, Atlas, Falcon, Ranger, Ambari, DataHub, Atlan, OpenMetadata, Collibra Data Intelligence Platform, Monte Carlo, Soda, Great Expectations, OpenLineage/Marquez, Delta Sharing. dbt: dbt Core/Cloud, dbt Semantic Layer. ________________________________________ Data Warehousing & BI Warehouses: Snowflake (Dynamic Tables, Snowpark, Snowpark Container Services, Native App Framework, Snowflake Cortex), Redshift, Azure Synapse, Databricks, BigQuery. BI/Viz: Tableau, Power BI, Microsoft Fabric, QlikView, QuickSight, Highcharts, Looker/Looker Studio, Mode, Superset, Metabase, ThoughtSpot, Cube. ________________________________________ Cloud Platforms & Multi-Cloud Engineering AWS (13+ yrs): Redshift, EMR, RDS/Aurora, S3, Glue, Lambda, Fargate, Kinesis, SageMaker, Bedrock, DynamoDB, API Gateway, CloudFormation, CDK, CodePipeline/CodeBuild/CodeDeploy, AWS Config, Secrets Manager, GuardDuty, Step Functions, Inferentia/Trainium. Azure (11+ yrs): ADF, Synapse, Microsoft Fabric, Azure ML Studio, Azure Databricks (Unity Catalog), ADLS Gen1/Gen2, Azure OpenAI, Event Hub, Stream Analytics, Microsoft AI Studio, MLflow, AKS. GCP (9+ yrs): GKE, BigQuery, Dataflow, Dataproc, Firestore, Bigtable, AI Platform/Vertex AI, TPUs, Cloud Run, Cloud Functions, Deployment Manager, Workload Identity, VPC Service Controls, Security Command Center (SCC), IAM. Search/Index: OpenSearch, Kendra. ________________________________________ Kubernetes, Containers & Microservices Clusters & ops: AKS/EKS/GKE, K3s, Rancher, Bare Metal; Helm, ArgoCD, FluxCD, Ingress, RBAC/Pod Security, image scanning, GitOps. Service mesh & networking: Istio, Linkerd, Cilium/eBPF; Knative, KEDA; messaging NATS JetStream; orchestration Temporal. ML on K8s: Kubeflow, Seldon, KServe. ________________________________________ DevOps, CI/CD & IaC Jenkins, GitHub Actions, GitLab CI/CD, CodePipeline/CodeBuild/CodeDeploy, Cloud Build, Azure DevOps; IaC Terraform, CloudFormation, ARM, CDK, Deployment Manager, Azure CLI, PowerShell; Packer, Crossplane, Terragrunt, Vault + External Secrets Operator, Backstage. ________________________________________ Search, Logging & Observability Elasticsearch (ELK), Apache Solr, Lucene, CloudSearch, OpenSearch (vector/ANN), Vespa, Meilisearch, Typesense, LanceDB, Vectara; Kibana, Grafana, Prometheus, Jaeger, OpenTelemetry, Filebeat, Tika, Rsync (real-time indexing, distributed search, analytics with Kafka). ________________________________________ Programming Languages & Full-Stack Python, Java (Maven), .NET (C#), Scala (SBT), Julia, Node.js, TypeScript/JavaScript, Angular, React.js, Spring/Spring Boot, FastAPI, gRPC, GraphQL, Next.js, NestJS, Prisma, Rust, Go, DBT Python APIs, Poetry, Encore Analytics Empower. ________________________________________ Enterprise Applications & Migrations SAP migrations (HANA, BW, BTP, CRM, ERP, EAM, PM) to AWS/Azure/GCP; VMware on AWS hybrid; legacy modernization; SAS Studio migrations; Solr Elasticsearch; SAP Datasphere, SAP Data Intelligence Cloud; ADF Mapping Data Flows, Informatica IICS, Talend. ________________________________________ High-Performance AI Infrastructure & AI Ops NVIDIA Mission Control, Bright Cluster Manager, Run:AI, DGX clusters; MIG, GPUDirect Storage, RDMA/InfiniBand; Slurm; NGC catalogs, NIM blueprints; Argo Workflows, Karpenter; GPU scheduling, multi-tenant isolation, air-gapped clusters, reproducibility, multi-cloud HPC. ________________________________________ Leadership & Management 17+ years leadership (incl. 4 years VP of Engineering) leading cross-functional teams, complex enterprise programs, solution architecture, and end-to-end delivery for Fortune 500 and public sector. Keywords: csharp continuous integration continuous deployment artificial intelligence machine learning javascript access management business intelligence sthree database sfour active directory rlang information technology golang business works hewlett packard fiveg mtv mountain view bay area California Texas Washington