Home

krishna Reddy - AI/ML engineer

[email protected]

Location: Frisco, Texas, USA

Relocation: Yes

Visa: H1B

Resume file: Krishna Resume_1776801486218.docx Please check the file(s) for viruses. Files are checked manually and then made available for download.

KRISHNA REDDY Senior AI/ML Engineer | Generative AI & Machine Learning +1 9408008619 | [email protected] PROFESSIONAL SUMMARY: Senior AI/ML Engineer with 8+ years of hands-on experience building and deploying production-grade Machine Learning and Generative AI systems across retail, healthcare, and banking domains. Designed and delivered enterprise scale Generative AI platforms using agentic architectures, Retrieval Augmented Generation (RAG), and rule governed workflows to support real world decision making. Strong experience translating business requirements into scalable AI/ML and Generative AI solutions in enterprise environments. Expertise in designing reusable ML pipelines, APIs, and production-grade AI systems with CI/CD Extensive experience working with GPT-4 / GPT-4 class models and LLaMA models, including LoRA based fine tuning for domain specific tasks such as product intelligence, clinical summarization, and content normalization. Strong expertise in LLM orchestration and agent workflows using LangGraph, LangChain, implementing stateful, multi-step reasoning, conditional branching, retries, guardrails, and controlled termination. Built and optimized RAG pipelines using Pinecone, FAISS and ChromaDB, incorporating metadata filtering, source allowlisting, freshness checks, and schema-constrained outputs to ensure grounded and auditable responses. Applied advanced prompt engineering techniques, including system prompts, few-shot learning, tool/function calling, and structured output validation to reduce hallucinations and enforce response consistency. Proven experience in LLMOps and MLOps, including model and prompt versioning, offline evaluation, drift detection, canary deployments, rollback strategies, and continuous monitoring. Developed and deployed low-latency inference services using FastAPI, Docker, and Kubernetes (AKS, EKS) to support real-time and near-real-time AI use cases. Hands-on experience with Vertex AI for model development and deployment, BigQuery for large-scale analytics, and GKE for containerized ML workloads. Strong cloud-native engineering experience across Microsoft Azure and AWS, leveraging Azure OpenAI Service, AKS, Azure Functions, API Management, AWS SageMaker, S3, EKS, EC2, and CloudWatch. Built scalable ML and data pipelines using Apache Spark, PySpark, Databricks, Kafka, Airflow, Delta Lake, and Snowflake for batch and streaming workloads. Implemented explainable and governed ML systems using SHAP, confidence scoring, audit logging, lineage tracking, and model risk management controls, supporting regulatory and compliance requirements. Hands on experience integrating structured and unstructured data via REST APIs, FHIR/HL7 interfaces, enterprise catalog systems, and real time inventory services. Domain experience in Retail, Healthcare, and BFSI (Banking) across large-scale enterprise programs TECHNICAL SKILLS Platform Tools & Technologies Languages & Scripting Python, SQL, PySpark, Bash, PowerShell ML & AI Frameworks Scikit learn, TensorFlow, PyTorch, XGBoost, SHAP Vector Databases Pinecone, ChromaDB, FIASS. Frameworks and Orchestration (Gen AI & LLM s) LangChain, LangGraph, LangFlow, AutoGen, RAG, Prompt Engineering. FastAPI, Redis API Gateway / Azure API Management Models and Fine-tuning (Gen AI & LLM s) GPT-3.5, GPT-4, OpenAI, LLaMA 3, LLM, BERT, Sentence Transformers, LoRA, PEFT, Hugging Face Transformers. MLOps & DevOps Tools MLOps / CI-CD: Docker, Kubernetes, Jenkins, GitHub Actions, Model Deployment, Monitoring, Drift Detection Cloud Platforms Microsoft Azure (Open AI, API, AKS, Functions), AWS (S3, Redshift, Lambda, RDS, CloudWatch), Google Cloud Platform (Vertex AI, Big Query, GKE, Dataflow) Data Engineering & ETL Databricks, Apache Kafka, Airflow, MongoDB, PostgreSQL, Delta Lake, SQL Server, APIs (Google Maps, USBR, USGS, SNOTEL) Visualization, productivity & collaboration Power BI, Tableau, Streamlet, VS Code, Git, Confluence, JIRA Domain Expertise Retail, Healthcare, Banking Analytics, Threat Detection, Regulatory Knowledge Bases, Claims & Underwriting, Forecasting, Risk & Fraud Analytics. PROFESSIONAL EXPERIENCE Client: Comerica Bank, Austin TX Nov 2024 Present Role: Senior AI/ML Engineer Generative AI Description: Designed and delivered a governed, agentic Generative AI decision support platform on Microsoft Azure, supporting frontline associates and customer service teams across retail locations. The system addressed fragmented and fast changing product data, complex policy interpretation, and real time inventory constraints to enable accurate, consistent, auditable decision making at enterprise scale. Technical Stack: Azure OpenAI (GPT-4 / GPT-4-class), LLaMA (LoRA fine-tuned) LangGraph, Pinecone, Embeddings, RAG, MLOps, LLMOps, Management API-Driven Systems, Microsoft Azure, leveraging Azure OpenAI Service, AKS, Azure Functions, Azure API Management, and Azure Monitor. Responsibilities: Implemented a deterministic, multi agent Generative AI architecture using Azure OpenAI (GPT-4), LangGraph, and rule-based orchestration, prioritizing repeatability, auditability, and operational safety over free form conversational behavior. Developed reusable AI components, APIs, and modular pipelines for scalable production deployment Translated business requirements into enterprise-scale Generative AI solutions for retail operations Built and maintained the agent orchestration layer with LangGraph, enabling stateful workflows, conditional branching, retries, guardrails, and controlled termination for multi-step reasoning. Developed a Product Intelligence Agent leveraging LLM reasoning, enterprise product catalog APIs, and vector similarity search with Pinecone to resolve SKUs and variants, manage discontinued products, and generate compatibility and substitution recommendations. Ensured secure AI deployment with RBAC, encrypted data pipelines, and enterprise compliance standards Validated and fine-tuned LLM models to improve accuracy, grounding, and response consistency Implemented a Policy Interpretation Agent combining GPT-4 based contextual reasoning with deterministic rule engines, configuration driven logic, and versioned policy artifacts to safely interpret return, warranty, and pricing policies. Built an Inventory & Availability Agent integrating real-time inventory APIs, location aware queries, and fulfillment business logic to produce store level, regional, and delivery aware recommendations. Defined microservices architecture, API contracts, and data flow pipelines Designed a Response Composition Agent using schema-constrained generation (JSON schemas structured outputs) to produce clear, step by step guidance optimized for frontline associate workflows. Engineered Retrieval-Augmented Generation (RAG) pipelines using Pinecone, metadata filtering, source allowlisting, and freshness validation to ground responses in authoritative, up to date enterprise data. Integrated open-source LLaMA models with LoRA fine tuning for domain specific tasks such as product attribute extraction, content normalization, and internal summarization, optimizing cost and latency for non-customer-facing workflows. Applied prompt engineering techniques, including system prompts, few-shot examples, tool/function calling, and output validation to minimize hallucinations and enforce response quality. Designed and deployed an end-to-end RAG-based document intelligence system using Vertex AI and Big Query for large-scale document processing and analytics Implemented MLOps / LLMOps pipelines for prompt versioning, model version pinning, offline evaluation, canary deployments, rollback strategies, and continuous improvement across agents. Integrated human-in-the-loop escalation workflows, implementing confidence thresholds, exception detection, and manual review paths for high-risk or policy exception scenarios. Built scalable data pipelines using Dataflow (Apache Beam) to ingest, preprocess, and transform unstructured documents (PDFs, text, logs) Developed end-to-end RAG pipelines including data ingestion, chunking, embedding generation, vector indexing, retrieval, and response generation Built high-performance REST APIs using FastAPI, enabling low-latency AI inference and seamless system integration Collaborated with platform, data, and compliance teams to design and implement end-to-end observability and audit logging, capturing agent execution paths, data lineage, policy versions, prompts, and final recommendations. Deployed and operated the platform on Microsoft Azure, leveraging Azure OpenAI Service, AKS, Azure Functions, Azure API Management, and Azure Monitor, optimizing latency, throughput, and inference cost. Implemented telemetry-driven evaluation, using logs, metrics, offline test cases, and error analysis to refine prompts, agent workflows, retrieval strategies, and model selection. Established model monitoring, evaluation, and CI/CD pipelines using Vertex AI Pipelines Client: GEICO, Dallas, TX Feb 2023 Oct 2024 Role: Machine Learning Engineer Description: Designed and delivered an AI-driven Clinical Decision Support System to proactively predict patient risks, support clinician decision making, and improve care outcomes at scale. The platform leveraged machine learning, deep learning, and NLP to analyze structured and unstructured clinical data while ensuring compliance with FDA AI/ML guidelines, and ethical AI standards. The system was architected for clinical interpretability, security, scalability, and continuous monitoring, enabling safe deployment in real-world environments. Technical Stack: Python, PyTorch, RAG, Hugging Face Transformers, Lora, BERT, PEFT, Pinecone, Retrieval Pipelines, AWS Textract, Tesseract, Fast API, REST APIs, DVC, AWS SageMaker, S3, EKS, EC2, IAM, CloudWatch, AWS CDK, Docker, Jenkins, GitHub, Git. Responsibilities: Developed predictive patient risk models (readmission risk, clinical deterioration, adverse events) using Python, PyTorch, XGBoost, and scikit-learn on longitudinal EHR data. Built end-to-end ML pipelines incorporating Apache Kafka for real-time ingestion of EHR updates, device telemetry, and lab results, and Apache Airflow for automated preprocessing, feature engineering, model training, and evaluation Validated model performance using AUC, F1, and domain-specific metrics to meet accuracy requirements Designed, trained, and optimized machine learning and deep learning models for clinical predictions Built LLM-powered clinical reasoning pipelines using Hugging Face Transformers and ClinicalBERT to summarize patient context and support clinician decision making. Implemented Retrieval-Augmented Generation (RAG) pipelines using vector embeddings and semantic search to ground model outputs in authoritative clinical guidelines and patient history. Integrated structured and unstructured clinical data via FHIR/HL7 APIs, ingesting labs, vitals, medications, diagnoses, and clinician notes. Implemented Responsible AI practices including fairness, bias detection, and governance frameworks Designed LLM evaluation frameworks measuring response quality, grounding, and hallucination rates Developed and deployed FastAPI-based inference APIs for real-time clinical predictions Designed explainable AI (XAI) frameworks using SHAP, attention mechanisms, and confidence scoring to ensure clinical interpretability and regulatory transparency. Implemented bias detection, fairness evaluation, and ethical AI safeguards using AIF360 and custom validation metrics, ensuring equitable performance across patient populations. Deployed secure, scalable inference services using Fast API, Docker, and Kubernetes (AWS EKS) to support low-latency clinical predictions in production. Built HIPAA compliant ML pipelines on AWS (SageMaker, EC2, S3, IAM) with encryption, access controls, and audit logging. Established model lifecycle management and governance using MLflow, DVC, and model registries, enabling traceability, reproducibility, and regulatory readiness. Implemented real-time monitoring, drift detection, and post-deployment surveillance using Prometheus and CloudWatch to ensure sustained clinical safety and performance. Automated CI/CD workflows using GitHub Actions and Jenkins, supporting continuous validation, deployment, and rollback of clinical ML models. Optimized API performance and scalability using Docker + Kubernetes (EKS) Collaborated with clinicians, compliance officers, and data teams to embed AI insights into existing clinical workflows without disrupting care delivery. Achieved 15%+ improvement in predictive model performance (AUC/F1) while maintaining clinical interpretability and reduced post deployment model incidents by 25% through continuous monitoring, drift detection, and governance controls. Client: S&P Global, India Dec 2019- Jul 2022 Role: Data Scientist Description: Worked as a Machine Learning Engineer supporting enterprise credit risk, fraud, and portfolio analytics within retail banking. Designed and operationalized end-to-end machine learning pipelines covering feature engineering, model development, validation, deployment, and monitoring using cloud-native and big-data technologies. Built scalable batch training and scoring workflows while adhering to model risk management (MRM), auditability, and regulatory compliance standards. Implemented model explainability, performance monitoring, and data drift detection to ensure long-term model stability. Collaborated with cross-functional risk, data science, and business teams to translate risk requirements into production-ready ML solutions. Technical Stack: Python, advanced SQL (Oracle, SQL Server), scikit-learn, XGBoost, PySpark, Apache Spark, Hive, Apache Airflow, MLflow, Docker, AWS (S3, EC2, EMR), SHAP, (PSI, CSI, AUC, KS), Git, Jenkins, Jira Responsibilities: Designed, developed, and productionized credit risk and delinquency prediction models using Python, scikit-learn, and XGBoost, supporting HSBC retail banking portfolio. Built scalable feature engineering pipelines using advanced SQL, PySpark, pandas, and Apache Spark, generating behavioral, transactional, and exposure-based features from high-volume financial data Worked in large-scale regulated enterprise programs aligned with model risk management (MRM) standards Created and maintained training, validation, and scoring datasets on cloud-based data platforms, enabling model development, periodic recalibration, and regulatory reporting. Implemented end-to-end ML pipelines covering data ingestion, feature generation, model training, validation, and batch scoring workflows. Deployed ML models as containerized services using Docker and automated workflows with Apache Airflow, improving reliability and repeatability of model execution. Established model lifecycle management and experiment tracking using MLflow, ensuring version control, reproducibility, and audit readiness. Developed model performance and stability monitoring frameworks, tracking AUC, KS, PSI, CSI, and data drift to ensure ongoing compliance with model risk management (MRM) standards. Produced model explainability and validation artifacts using SHAP, supporting internal model validation, audits, and regulatory reviews. Optimized Spark and cloud workloads (AWS S3, EC2, EMR) to improve execution performance and control infrastructure costs. Strengthened CI/CD and release management processes using Git and Jenkins, enabling controlled model deployments and change tracking. Enforced secure data access, encryption, and audit logging, aligning ML solutions with enterprise security, risk, and compliance requirements. Collaborated with risk analysts, data scientists, and business stakeholders to refine features, decision thresholds, and risk strategies across the credit lifecycle. Client: Flutter Entertainment, India Aug 2017 to Nov 2019 Role: Data Engineer Description: As a Data Engineer at Flutter Entertainment designed, developed, and maintained scalable data integration solutions supporting clinical, operational, and reporting needs. I built and optimized robust ETL pipelines to ingest, cleanse, and transform large volumes of structured and semi-structured data, ensuring data accuracy, consistency, and timely availability for analytics, regulatory reporting, and business intelligence. I collaborated closely with business analysts, data architects, QA, and domain stakeholders to deliver reliable, compliant, and high-quality data solutions aligned with industry standards. Technical Stack: AWS (S3, Glue, Redshift), Apache Airflow, Informatica PowerCenter, Talend, Python, advanced SQL (Oracle, SQL Server, MySQL), Jira, Git, REST APIs, Microsoft Excel. Responsibilities: Developed and optimized ETL pipelines using Informatica PowerCenter, Talend, and custom Python/SQL scripts to process transactional, clinical, claims, and reference data from multiple source systems. Led the migration of legacy on prem ETL workflows to AWS, rearchitecting pipelines using S3, Glue, and Redshift to improve scalability, performance, and cost efficiency. Implemented advanced SQL transformations, aggregations, and reconciliations to ensure accurate consolidation of patient, provider, and operational datasets for dashboards and reporting. Designed and documented star and snowflake schemas to support analytics, KPI reporting, and self-service BI use cases. Implemented data quality validation and reconciliation checks, proactively identifying and resolving data inconsistencies to meet compliance and reporting requirements. Collaborated with data governance and QA teams to support metadata management, lineage tracking, and audit readiness. Provided production support, root cause analysis, and performance tuning for ETL pipelines, actively participating in UAT cycles and production deployments to ensure minimal downtime. CERTIFICATIONS Microsoft Certified: Azure Data Scientist Associate Microsoft Certified: Azure AI Engineer Associate AWS Certified Machine Learning Specialty EDUCATION Master of Science Artificial Intelligence University of North Texas Keywords: continuous integration continuous deployment quality analyst artificial intelligence machine learning business intelligence sthree Kansas Texas