Home

Abhinash Reddy Peddyreddy - AI ML Engineer

[email protected]

Location: Chicago, Illinois, USA

Relocation: yes

Visa: GC

Resume file: Abhinash_AI_ML_Engineer_1774272758317.docx Please check the file(s) for viruses. Files are checked manually and then made available for download.

Abhinash Reddy Peddyreddy Sr AI/ML Engineer [email protected] (270) 226-0303 Professional Summary Sr AI/ML Engineer with 11+ years of experience architecting and delivering enterprise-grade AI, ML, and Generative AI solutions across Healthcare, Telecom, and FinTech domains. Expertise in designing Agentic AI Workflows and autonomous multi-agent systems using frameworks like LangGraph, CrewAI, and AutoGen for complex, multi-step business processes. Production-scale deployment of RAG and GraphRAG architectures, utilizing Vector Databases (Pinecone, Milvus, Weaviate) and Knowledge Graphs (Neo4j) to minimize hallucinations and maximize retrieval accuracy. Full-stack proficiency across major AI Clouds, including AWS Bedrock, Azure OpenAI (Prompt Flow), and Google Vertex AI, ensuring vendor-agnostic deployment and high availability. Advanced LLM Fine-tuning and Optimization, specializing in PEFT, LoRA, and QLoRA techniques to adapt open-source models (Llama 3.1, Mistral, Gemma) for domain-specific tasks. Strategic Prompt Engineering leadership, implementing Chain-of-Thought (CoT), ReAct, and Tree-of-Thought prompting to drive reasoning capabilities in customer-facing applications. Deep implementation of LLMOps and Observability, utilizing LangSmith, Arize Phoenix, and TruLens for model evaluation, cost tracking, and real-time drift detection. Architected scalable inference pipelines using vLLM, NVIDIA NIM, and Triton Inference Server to optimize latency, throughput, and GPU utilization. Expertise in the Apache Spark ecosystem (PySpark, MLlib) and Databricks for processing petabyte-scale datasets and building robust feature stores for AI model training. Proven success in Computer Vision and Multimodal AI, including handwritten OCR, object detection (YOLO), and image segmentation pipelines integrated with LLM reasoning. Strong background in AI Governance and Security, implementing guardrails (NeMo, Llama Guard) for PII detection, content filtering, and protection against prompt injection attacks. Developed sophisticated predictive and decision science models using XGBoost, LightGBM, and Ensemble methods for financial risk assessment and high-frequency forecasting. Skilled in software internationalization and globalization, ensuring multilingual LLM readiness and translation validation for global software deployments. Experienced in CI/CD for AI/ML lifecycle, managing end-to-end productionization through Docker, Kubernetes (K8s), and Infrastructure as Code (Terraform/CDK). Quantifiable track record of business impact, including reducing operational costs by up to 40% through automation and improving model accuracy by 25%+ in mission-critical environments. Technical Skills: Programming Languages Python (Expert), SQL, Java, C, C++, R, TypeScript, JavaScript, Next.js, ReactJS, Scala, Impala, Hive, Shell Scripting (Bash) Statistical Methods Statistical Inference, Hypothesis Testing, Experimental Design (A/B Testing), Multivariate Analysis, Time Series Analysis, Auto-correlation, Bayesian Statistics (Bayes' Theorem), Statistical Modelling, ANOVA, Chi-Square Tests, Correlation and Covariance Analysis, Probability Distributions, Sampling Techniques, Residual Analysis, Cross-Validation, Descriptive Statistics. Machine Learning XGBoost, LightGBM, CatBoost, Ensemble Methods, Random Forest, Support Vector Machines (SVM), Recommendation Systems, Dimensionality Reduction (PCA, t-SNE), Feature Engineering, Hyperparameter Optimization (Optuna, Ray Tune), Time Series Forecasting (Prophet, ARIMA, SARIMA, LSTM), Model Interpretability (SHAP, LIME), Statistical Inference, Hypothesis Testing, A/B Testing, Scikit-learn, Azure Machine Learning Services, Model Evaluation Metrics (ROC-AUC, F1-Score, Precision-Recall), Error Analysis, Bias-Variance Tradeoff, Scalable ML Pipeline Design. Deep Learning PyTorch, TensorFlow, Keras, Deep Neural Networks (DNN), Convolutional Neural Networks (CNN), Transformer Architectures, Attention Mechanisms, Transfer Learning (ResNet, Inception, MobileNet), Autoencoders, Object Detection (YOLO, Faster R-CNN), Image Segmentation (U-Net, Mask R-CNN), OCR, Optimization Algorithms (Adam, SGD, RMSprop), Regularization Techniques (Dropout, Batch Normalization), Distributed Model Training, Model Quantization, Model Deployment. Natural Language Processing (NLP) Transformer Models (BERT, GPT, T5), Hugging Face, SpaCy, NLTK, Large Language Modeling, Sequence-to-Sequence Models, Attention Mechanisms, Encoder-Decoder Architectures, Word Embeddings (Word2Vec, FastText), Named Entity Recognition (NER), Sentiment Analysis, Text Summarization, Machine Translation, Question Answering Systems, Topic Modeling (LDA), Dependency Parsing, Sequence Labeling, Semantic Search, Word and Sentence Similarity, Language Model Pre-training and Fine-tuning. Generative AI LLMs (GPT-4, Llama 3.1, Claude 3.5, Mistral, Gemma), Multi-Agent Systems (LangGraph, CrewAI, AutoGen), Retrieval-Augmented Generation (RAG), GraphRAG, Knowledge Graphs (Neo4j), Vector Databases (Pinecone, Milvus, Weaviate, Chroma, Faiss), LLM Orchestration (LangChain, LlamaIndex, Haystack), LangSmith, Fine-tuning (LoRA, QLoRA, PEFT), Prompt Engineering (CoT, ReAct), AI Guardrails (NeMo, Llama Guard), Local LLM Deployment (Ollama, vLLM, TGI), Inference APIs (Groq, NVIDIA NIM), Document Chunking & Ingestion Pipelines, Diffusion Models, Multimodal AI (CLIP, DALL-E), Transformer Architectures (T5, BERT), GANs (StyleGAN, CycleGAN), Variational Autoencoders (VAEs), Reinforcement Learning (RLHF), Zero-Shot & Self-Supervised Learning, AWS Bedrock. MLOps & LLMOps CI/CD for ML/LLM Lifecycle, Model Deployment (REST API, gRPC, Batch), Scalable LLM Serving (vLLM, TGI, NVIDIA NIM), LLM Observability & Evaluation (LangSmith, TruLens, Ragas, PromptLayer), Experiment Tracking (MLflow, Weights & Biases), Model Registry & Versioning, Data Versioning (DVC), Feature Stores (Feast), Orchestration (Kubernetes, Airflow, Kubeflow, Argo Workflows), Infrastructure as Code (Terraform, CloudFormation, CDK), Containerization (Docker), Monitoring & Alerting (Prometheus, Grafana, Drift Detection), Token Usage & Cost Tracking, Cloud AI Platforms (AWS SageMaker, AWS Bedrock, Google Vertex AI, Azure AI Services), Azure DevOps, GitHub Actions. Big Data: Apache Spark, Databricks, Scala, Hadoop, Hive, HBase, Amazon Kinesis, Azure Synapse Analytics, Azure Data Lake Storage (ADLS), Azure Blob Storage, AWS S3, Delta Lake, Real-time Data Ingestion, Distributed Computing, Large-scale Data Processing. Amazon Web Services: AWS Bedrock, SageMaker, EMR, MSK, Kinesis, Glue, Athena, S3, DynamoDB, Lambda, EC2, Step Functions, API Gateway, AWS CDK, CloudFormation, IAM, CloudWatch, Glacier, Lex, Rekognition, Transcribe, QuickSight, CodeCommit. Database Servers: PostgreSQL, MySQL, Microsoft SQL Server, Amazon Redshift, Amazon RDS, MongoDB, Teradata, SQLite, NoSQL, Relational Database Design, Query Optimization. Other Tools & Technologies: Git, GitHub, GitLab, Docker Compose, CUDA Toolkit, Linux/Unix Command Line, Bash Scripting, Jenkins, Terraform, Nginx, Redis, REST APIs, Swagger, Postman, Conda, Virtualenv, Makefile, YAML, JSON, VS Code, Jupyter Notebook, Google Colab, Power BI, Tableau. Professional Experience: Client: Oracle, Austin, Texas Dec 2023 Present Role: Sr AI/ML Engineer Project: Oracle AI Cloud Services Intelligent Automation and Generative AI Platform Responsibilities: Led the architectural design and deployment of Generative AI solutions within Oracle Cloud Infrastructure (OCI), integrating LLMs to drive intelligent automation and system diagnostics across global enterprise applications. Architected production-grade Retrieval-Augmented Generation (RAG) frameworks leveraging Oracle Database 23ai Vector Search, LangChain, and FAISS to revolutionize enterprise search capabilities. Architected production-grade Retrieval-Augmented Generation (RAG) frameworks leveraging Oracle AI Vector Search, LangChain, and FAISS to revolutionize enterprise search capabilities within Oracle Fusion Cloud. Engineered sophisticated multi-agent autonomous systems using LangChain and Vertex AI to automate high-value workflows, including service ticket classification and proactive IT support curation. Orchestrated the fine-tuning and optimization of state-of-the-art LLMs (GPT-4, Claude, Llama-3) on OCI GPU clusters, achieving significant gains in latency reduction and domain-specific response fidelity. Developed high-performance NLP pipelines using BERT and Hugging Face Transformers for intelligent document processing and knowledge extraction from complex multi-source support datasets. Scaled AI services through microservice architectures using FastAPI and Flask, delivering RESTful APIs for real-time anomaly detection and predictive forecasting integrated with Oracle Digital Assistant. Established comprehensive LLMOps and MLOps ecosystems using MLflow and Airflow, streamlining the end-to-end lifecycle from training to continuous governance across multi-cloud environments. Deployed high-impact predictive models (XGBoost, PyTorch) for capacity planning and system health, resulting in a documented 25% reduction in mean time to resolution (MTTR). Built petabyte-scale data ingestion and monitoring frameworks utilizing Kafka, Spark Streaming, and Delta Lake to process billions of telemetry records and log events daily. Strategized the implementation of enterprise feature stores using PySpark and Snowflake, enabling scalable, consistent data access for model training across distributed teams. Pioneered LLM evaluation and observability standards using TruLens and PromptLayer to ensure model explainability, bias mitigation, and alignment with corporate Responsible AI frameworks. Secured AI infrastructure by integrating with OCI Identity and Access Management (IAM) and Vault services, ensuring robust data protection and secure key management for sensitive cloud operations. Automated global AI infrastructure deployment using Infrastructure as Code (Terraform, Ansible), ensuring consistency across development, staging, and production environments. Championed the GenAI Center of Excellence, defining enterprise governance standards, deployment templates, and mentoring cross-functional teams on modern AI integration patterns. Delivered end-to-end AI/ML solutions across Azure and AWS ecosystems, leveraging Azure ML Studio, Bedrock, and Synapse to build scalable knowledge management and RAG-based search systems. Environment: Python, PySpark, TensorFlow, PyTorch, LangChain, Hugging Face, Oracle Cloud (OCI), Vertex AI, Azure ML, AWS Bedrock, FastAPI, MLflow, Airflow, Kubernetes (OKE/EKS), Terraform, Snowflake, Kafka, Delta Lake. Client: Paypal, CA(Remote) Apr 2021 Dec 2023 Role: AI/ML Engineer Project: Generative AI Driven Fraud Intelligence and Customer Experience Automation Platform Responsibilities: Architected the transition from traditional discriminative ML to Generative AI systems, leveraging LangChain and Vertex AI to automate fraud analytics and conversational support. Engineered production-grade Retrieval-Augmented Generation (RAG) pipelines, integrating enterprise data lakes with vector databases (Pinecone, FAISS) to enable semantic search for fraud investigation. Developed autonomous multi-agent systems using LangChain and early-access Google Agent tools to automate compliance workflows, including KYC/AML validations and regulatory reviews. Optimized fraud detection accuracy by 26% through the development of real-time ML pipelines using XGBoost, LightGBM, and PyTorch, achieving sub-second inference for millions of daily events. Led LLM benchmarking and fine-tuning initiatives (GPT-4, Llama 2, and Claude) to optimize precision and cost-effectiveness across unstructured financial datasets. Scaled NLP architectures for intent detection and sentiment analysis using Hugging Face Transformers, reducing manual triage time for customer disputes by 38%. Designed high-throughput streaming frameworks using Kafka, Spark Streaming, and Delta Lake to process real-time payment signals at petabyte scale. Standardized LLMOps and MLOps lifecycles using MLflow and Airflow, implementing automated versioning, CI/CD, and retraining triggers based on model drift. Pioneered Graph-based risk intelligence by building link-analysis models with Neo4j and GraphSAGE to identify and neutralize fraudulent transaction networks in real time. Established AI Governance frameworks using RLHF (Reinforcement Learning from Human Feedback) and TruLens to monitor for bias and maintain PCI-DSS/GDPR compliance. Built scalable AI microservices via FastAPI, exposing high-availability RESTful APIs for real-time transaction scoring and generative assistant responses. Reduced compute costs by 30% by optimizing data ingestion and feature engineering pipelines within Snowflake, Databricks, and PySpark. Automated multi-cloud AI infrastructure using Terraform, ensuring secure and elastic deployment across AWS and Azure environments. Integrated enterprise-scale Knowledge Bases within AWS (S3, Lambda, Glue) to provide context-aware intelligence for risk investigation units. Mentored cross-functional teams on LLM fine-tuning, agent orchestration, and modern MLOps best practices to foster a culture of AI excellence. Environment: Python, PySpark, TensorFlow, PyTorch, LangChain, Hugging Face, Vertex AI, AWS Bedrock/SageMaker, Azure ML, FastAPI, MLflow, Airflow, Docker, Kubernetes (EKS), Terraform, Snowflake, Kafka, Delta Lake, Neo4j, Pinecone. Client: Verizon Communications, Irving, Texas Nov 2019 Mar 2021 Role: Machine Learning Engineer Project: AI-Driven Network Optimization & Customer Retention Platform Responsibilities: Developed high-precision predictive models using XGBoost, LSTM, and Prophet to forecast network traffic and subscriber churn with a documented 92% accuracy rate. Implemented real-time anomaly detection systems leveraging Kafka Streams, Spark Structured Streaming, and PyTorch, successfully reducing false-positive alarms by 28%. Integrated complex telemetry and IoT sensor data from global tower and router networks into centralized data lakes for advanced time-series forecasting and fault diagnostics. Designed NLP-based analytics pipelines using BERT and spaCy to extract sentiment and intent from customer feedback, call transcripts, and internal NOC complaint logs. Engineered high-performance RESTful inference APIs using FastAPI, embedding real-time predictions directly into mission-critical network operations dashboards. Containerized and orchestrated ML workloads using Docker and Kubernetes (Amazon EKS), implementing horizontal pod autoscaling to handle fluctuating global traffic volumes. Standardized MLOps workflows with MLflow and Airflow, automating experiment tracking, model versioning, and CI/CD deployments to ensure rapid iteration cycles. Led the implementation of enterprise feature stores and distributed transformations using PySpark and AWS Glue to process and model petabyte-scale datasets. Optimized deep learning inference by 40% through model quantization and ONNX Runtime, partnering with cross-functional teams to meet strict SLA latency requirements. Deployed Graph Neural Network (GNN) prototypes on Neo4j and Amazon Neptune for sophisticated root-cause analysis of interdependent network node failures. Operationalized model drift detection and automated retraining triggers using statistical monitoring and CloudWatch, maintaining long-term model precision. Improved Mean Time to Detect (MTTD) network faults by 19% through A/B testing and simulation experiments of model-driven maintenance strategies. Integrated AI-driven alerting into enterprise tools like ServiceNow and Grafana, providing field engineers with real-time, actionable insights into network health. Environment: Python, PySpark, TensorFlow, PyTorch, XGBoost, MLflow, FastAPI, Kubernetes (EKS), Kafka, Spark Streaming, Airflow, AWS SageMaker/Lambda/Glue, GCP Vertex AI, ONNX, Neo4j, Grafana, Power BI, BERT, spaCy. Client: UnitedHealth Group (UHG), North Carolina, USA Mar 2016 Oct 2019 Role: Data Analyst Project: Predictive Health Risk Stratification & Cost Optimization Platform Responsibilities: Developed predictive statistical models using XGBoost, Random Forest, and Logistic Regression to forecast patient readmissions, high-cost claimants, and chronic disease progression. Enhanced model performance through rigorous feature selection, correlation analysis, and k-fold cross-validation, achieving a documented 25% improvement in prediction precision. Automated data preparation and analytical workflows using Python and SQL, leveraging AWS (S3, EC2) to handle large-scale healthcare datasets for risk stratification. Utilized early AWS SageMaker features (post-2017) to streamline model training and validation processes for scalable patient risk scoring. Implemented NLP-based text processing using NLTK and spaCy to identify ICD/CPT codes and clinical conditions from unstructured physician notes and EHR narratives. Prototyped sequential patient journey models using TensorFlow 1.x and Keras to analyze time-series risk progression and intervention timing. Partnered with data engineering teams to transition legacy SAS/SQL analytics workloads into Python-based cloud architectures on AWS and Databricks. Designed and deployed executive-level dashboards in Tableau and Power BI to visualize patient risk trends, cost metrics, and the effectiveness of care programs. Collaborated closely with clinical and actuarial teams to translate complex model outputs into actionable outreach strategies and preventive care campaigns. Applied model interpretability techniques (SHAP, LIME) and internal validation frameworks to ensure model transparency and clinical fairness. Maintained strict adherence to HIPAA and PHI data governance standards, documenting data lineage and model logic for regulatory audit compliance. Conducted longitudinal A/B testing to compare ML-driven patient outreach against traditional rule-based triggers, resulting in a 17% increase in early-risk detection. Mentored junior analysts in statistical evaluation, feature engineering, and the use of Jupyter Notebooks for reproducible research. Presented data-driven insights and model performance metrics to clinical leadership, facilitating the adoption of predictive programs across the enterprise. Environment: Python (3.x), SQL, Pandas, NumPy, Scikit-learn, TensorFlow 1.x, Keras, XGBoost, NLTK, spaCy, AWS (S3, EC2, SageMaker), Tableau, Power BI, Jupyter Notebook, Git, HIPAA Compliance. Client: JP Morgan Chase, Jersey City, NJ December 2013 March 2016 Role: Python Developer Project: Enterprise Risk & Customer Analytics Platform Responsibilities: Developed Python-based automation scripts for large-scale data extraction and transformation (ETL) from internal transactional databases, improving processing efficiency by 30%. Built and maintained data ingestion pipelines to consolidate credit card and loan data into on-premise data warehouses using Python (Requests, Pandas) and SQL. Designed and optimized relational data models (Star and Snowflake schemas) in Teradata to support marketing and compliance analytics reporting. Created automated data validation frameworks in Python to ensure data consistency for regulatory reporting (CCAR/Basel III datasets). Utilized shell scripting and Cron to orchestrate daily data refresh cycles and monitor pipeline dependencies. Collaborated with Risk Analytics teams to assist in preparing datasets for predictive modeling, focusing on feature extraction for credit default and delinquency analysis. Implemented statistical analysis scripts using Scikit-learn and NumPy to identify customer spending patterns and segment behavior for business teams. Developed complex SQL procedures and materialized views in Teradata and Oracle to support downstream risk reporting dashboards. Built internal reporting tools and visualizations using Tableau and Matplotlib to track delinquency trends and fraud risk indicators. Authored data quality and lineage documentation to ensure adherence to internal data governance and audit standards. Contributed to the optimization of legacy Perl/Shell scripts by migrating them to Python 2.7/3.4, resulting in a 25% reduction in maintenance overhead. Participated in A/B testing cycles for marketing campaigns by managing data splits and calculating performance metrics. Environment: Python (2.7/3.4), SQL, Teradata, Oracle, Shell Scripting, Linux (RHEL), Tableau, Pandas, NumPy, Scikit-learn, Matplotlib, SVN/Git. Keywords: cprogramm cplusplus continuous integration continuous deployment artificial intelligence machine learning javascript business intelligence sthree rlang information technology California New Jersey