Home

Achyutha P - Senior Gen AI/ML Engineer

[email protected]

Location: Jersey City, New Jersey, USA

Relocation: YES

Visa: GC

Resume file: AchyuthaP_Gen_AI_ML_Engineer_Resume_1771944017215.docx Please check the file(s) for viruses. Files are checked manually and then made available for download.

Name: ACHYUTHA P Role: Generative AI & Machine Learning Engineer Email address: [email protected] LinkedIn: http://www.linkedin.com/in/achyutha-p-89a3b5216 Contact: +1 5513614995 PROFESSIONAL SUMMARY Senior Generative AI & Machine Learning Engineer with 11+ years of hands-on experience building and deploying production-grade AI systems across finance, healthcare, retail, insurance and telecommunications. Specialized in LLM-powered applications, Retrieval-Augmented Generation (RAG), model fine-tuning, prompt engineering and scalable cloud-based ML deployments. Built and deployed enterprise-grade Generative AI applications using Amazon Bedrock and Azure OpenAI, LangChain and LangGraph, implementing Retrieval-Augmented Generation (RAG) pipelines for document intelligence and knowledge retrieval. Engineered multi-agent LLM workflows and RAG pipelines integrating Amazon OpenSearch and Azure Cognitive Search, FAISS and Pinecone for domain-specific financial and healthcare use cases. Developed machine learning pipelines using AWS SageMaker, Vertex AI, Azure ML, Databricks and MLflow to support model training, evaluation, deployment and monitoring in cloud environments. Developed and fine-tuned transformer-based NLP models using BERT, GPT, LLaMA, Mistral and PEFT techniques (LoRA/QLoRA) for financial document analysis, clinical text processing, sentiment analysis and entity extraction in high-stakes, regulated environments. Applied time-series forecasting, ensemble learning and deep learning models using TensorFlow, PyTorch, XGBoost, ARIMA, Prophet and LSTM to support demand prediction, risk assessment, fraud detection and pricing optimization across large-scale production environments. Developed scalable data pipelines using PySpark, AWS Glue, Google Cloud Dataflow, Azure Data Factory, Delta Lake, Snowflake and dbt to process structured and unstructured data for ML model training and real-time inference. Integrated explainability techniques using SHAP and LIME to enhance model transparency and meet compliance requirements in regulated environments. Deployed production ML and GenAI services using Docker, Amazon EKS, GKE and Azure Kubernetes Service (AKS), supporting scalable real-time inference and monitoring using Amazon CloudWatch, Google Cloud Monitoring, Azure Monitor and Prometheus. Applied knowledge graph techniques using Neo4j and NetworkX to enhance fraud detection and compliance analytics use cases. Built modular LLM application layers integrating retrieval, prompt orchestration, model inference and response validation to deliver scalable GenAI services. Collaborated with data engineering, DevOps and product teams to deliver production-ready AI and Generative AI solutions aligned with business and compliance requirements. TECHNICAL SKILLS Programming & Core Technologies Python, SQL, PySpark, SAS, Java, Bash Generative AI & LLM Platforms Amazon Bedrock (Claude/Titan), LangChain, LangGraph, Semantic Kernel, Hugging Face Transformers, PEFT (LoRA, QLoRA), MedLM, RAG Architectures, Azure OpenAI (GPT-4) Vector Databases & Search FAISS, ChromaDB, Pinecone, Amazon OpenSearch, Azure Cognitive Search, Semantic Search, Hybrid Retrieval (BM25 + Vector) AI/ML Frameworks & Libraries PyTorch, TensorFlow, scikit-learn, XGBoost, Random Forest, Gradient Boosting Cloud Platforms & AI/ML Services AWS (core production experience): SageMaker, EMR, Redshift, EKS, OpenSearch, Comprehend, Textract, CloudWatch, Lambda, EC2 GCP (HCA projects): Vertex AI, BigQuery, Cloud Dataflow, Dataproc, GKE, Cloud Build, Document AI, Cloud Healthcare API, Cloud Monitoring, Looker Studio Azure: Azure ML, Azure Databricks, Azure Synapse Analytics, Azure Data Factory, Azure Functions, AKS, Azure Monitor, Azure DevOps Data warehousing / processing: Snowflake, Delta Lake MLOps & DevOps MLflow, Docker, Kubernetes (AKS, EKS, GKE), Azure DevOps, Git, Jenkins, CI/CD Pipelines, Model Monitoring, Drift Detection NLP & Transformers BERT, GPT, LLaMA, Mistral, spaCy, Text Summarization, Sentiment Analysis, Entity Recognition, Clinical NLP Time-Series & Forecasting ARIMA, Prophet, LSTM Graph Databases & Knowledge Graphs Neo4j, NetworkX, Graph Embeddings, Relationship Modeling Big Data & ETL PySpark, pandas, AWS Glue, Google Cloud Dataflow, Delta Lake, dbt, Alteryx, Apache Airflow, Azure Data Factory Data Warehousing & Databases Amazon Redshift, Snowflake, BigQuery, SQL Server, MySQL, Oracle, Query Optimization, Data Modeling Data Visualization & Business Intelligence Power BI, Tableau, Amazon QuickSight, Matplotlib, Seaborn, Looker Studio Responsible AI & Explainability SHAP, LIME, Amazon Bedrock Guardrails, Azure AI Content Safety Statistical Analysis & Experimentation A/B Testing, Hypothesis Testing, Cohort Analysis, Customer Segmentation, Statistical Modeling, GLMs Monitoring & Observability Prometheus, AWS CloudWatch, Google Cloud Monitoring, Azure Monitor, Application Insights APIs & Deployment FastAPI, OAuth2, RBAC, REST APIs, Real-Time Inference, Batch Scoring, Event-Driven Architectures WORK EXPERIENCE Client: Jefferies Financial Group Inc, New York, NY May 2024 Present Role: GenAI/ML Engineer Built and deployed AI/ML and Generative AI applications on AWS using Amazon SageMaker, Amazon EMR and Amazon Redshift to support secure and scalable model training and inference for financial research and risk analytics use cases. Developed ML pipelines using Python, Amazon SageMaker and MLflow to support reproducible experimentation, CI/CD integration and production deployment. Deployed LLM-powered enterprise assistants using Amazon Bedrock (Claude / Titan) and LangChain to automate financial document analysis, research summarization and knowledge retrieval, significantly reducing manual search and review effort for analysts. Engineered multi-agent Generative AI workflows using LangGraph and LangChain Agents to automate ingestion, retrieval and validation of SEC filings and financial research reports, improving workflow efficiency by ~33%. Developed Retrieval-Augmented Generation pipelines using Amazon OpenSearch with hybrid BM25 and vector retrieval, along with FAISS and Pinecone, implementing hierarchical chunking, semantic re-ranking and contextual filtering to reduce hallucinations by ~28% and improve citation accuracy. Fine-tuned domain-adapted LLMs using PEFT (LoRA/QLoRA) with PyTorch and Hugging Face Transformers, improving financial summarization and risk entity extraction accuracy by ~14% on internal evaluation benchmarks. Implemented query classification and routing logic to dynamically direct complex financial queries to Amazon Bedrock (Claude 3) and route routine requests to optimized open-source models, reducing average inference cost by ~20%. Built LLM-powered analyst copilots supporting earnings analysis, compliance workflows and policy navigation using LangChain, Amazon Bedrock and vector search, accelerating onboarding and improving analyst productivity. Deployed ML and GenAI workloads using Amazon SageMaker and Amazon EKS, supporting monitoring, retraining workflows and production inference. Developed secure, production-grade GenAI APIs using FastAPI, OAuth2 and RBAC, enabling seamless integration into internal financial systems. Implemented real-time inference pipelines using event-driven architectures, reducing prediction latency by ~35% for trading and risk analysis workflows. Integrated knowledge graph modeling using Neo4j to enhance fraud detection and compliance analytics through relationship-aware risk scoring. Applied Responsible AI practices using Amazon Bedrock Guardrails, SHAP and LIME to improve explainability, auditability and regulatory compliance of AI-generated financial outputs. Built and ran LLM evaluation pipelines using automated metrics and human-in-the-loop validation to compare Amazon Bedrock foundation models and fine-tuned open-source models. Built production monitoring and observability systems using Amazon CloudWatch, AWS X-Ray, Prometheus, reducing MTTR by ~30% and improving reliability of AI services. Technologies: Amazon Bedrock (Claude 3/Titan), LangChain, LangGraph, Amazon OpenSearch, FAISS, Pinecone, Hugging Face Transformers, PyTorch, PEFT (LoRA/QLoRA), Python, FastAPI, Docker, Amazon EKS, Amazon SageMaker, Amazon EMR, Amazon Redshift, Amazon Comprehend, AWS Textract, PySpark, MLflow, Amazon CloudWatch, AWS X-Ray, Prometheus, SHAP, Amazon Bedrock Guardrails, Neo4j________________________________________ Client: HCA Healthcare Inc., Nashville, TN November 2022 April 2024 Role: AI/ML Engineer & Data Scientist Developed and deployed production-grade ML models for patient risk prediction and stratification using Python, XGBoost, Random Forest and Google Cloud Vertex AI, improving early-risk identification and supporting proactive care delivery initiatives. Developed ML pipelines using Vertex AI, MLflow and Cloud Build (CI/CD) to reduce model release cycle time by ~25%. Built scalable data ingestion and ETL pipelines using Cloud Dataflow, Google Cloud Storage and BigQuery, improving data availability and enabling faster iteration for downstream modeling. Engineered claims fraud and anomaly detection models using ensemble learning techniques, reducing false positives by approximately 30% and improving investigation efficiency. Developed distributed data processing pipelines using PySpark, SQL and Google Dataproc improving large-scale analytics performance by ~35% over legacy systems. Applied deep learning models using TensorFlow and PyTorch for classification, anomaly detection and forecasting across clinical datasets, improving model robustness across retraining cycles. Deployed low-latency inference endpoints using Vertex AI, reducing prediction response time for patient risk scoring workflows. Implemented explainable AI (XAI) frameworks using SHAP and LIME, reducing clinician review time by ~20% while maintaining HIPAA-compliant audit transparency. Integrated external social determinant datasets through advanced feature engineering and data fusion, enhancing patient outcome modeling and risk stratification accuracy. Built model monitoring and drift detection dashboards using Looker Studio and Google Cloud Monitoring to track performance degradation and trigger retraining workflows. Developed NLP pipelines for clinical note analysis using BERT, spaCy and Hugging Face Transformers, improving structured data extraction accuracy. Implemented OCR and document intelligence workflows using Google Cloud Document AI and Vision API, reducing manual claims review effort. Containerized and deployed ML services using Docker and Google Kubernetes Engine (GKE) to support scalable, reliable inference across clinical systems. Developed GenAI-powered clinical documentation workflows using Vertex AI and MedLM (Med-PaLM 2) on Google Cloud, supporting real-time ambient medical note generation from clinician-patient conversations across emergency department sites, reducing physician documentation time and integrating with EHR systems. Technologies: Python, SQL, PySpark, XGBoost, Random Forest, TensorFlow, PyTorch, BERT, Hugging Face Transformers, spaCy, Google Cloud (Vertex AI, BigQuery, Cloud Dataflow, Dataproc, Cloud Storage, Document AI, Vision API, Cloud Healthcare API, GKE, Cloud Build, Cloud Monitoring), MLflow, Docker, Looker Studio, SHAP, LIME, LangChain, Vertex AI (Med-PaLM 2) ________________________________________ Client: Target Corp, Minneapolis, MN January 2019 - October 2022 Role: AI/ML Engineer Developed scalable ML-ready data pipelines using Python, PySpark and Azure services to support demand forecasting and pricing optimization. Built advanced feature engineering frameworks using SQL, dbt and Snowflake to generate time-series, customer-level and product-level features for downstream ML model training. Developed and deployed demand forecasting models using ARIMA, Prophet, LSTM, TensorFlow and PyTorch, improving forecast accuracy by ~25% on high-volume and seasonal SKUs. Implemented robust model evaluation and backtesting frameworks using RMSE, MAPE and rolling-window validation, ensuring statistical models and deep learning models met production performance thresholds. Applied ensemble learning techniques using XGBoost and Random Forest to optimize pricing, promotion planning and markdown strategies across retail categories. Integrated external market signals (Nielsen data, promotional calendars, competitive pricing, social sentiment) into model features to improve responsiveness to seasonality and promotional demand shifts. Operationalized ML models using Azure Functions and CI/CD pipelines, enabling consistent batch and scheduled inference across 1,800+ retail locations. Developed optimized ETL workflows using Azure Data Factory, Alteryx and Python scripting to automate data preparation and reduce manual processing effort by ~40%. Implemented data and model governance controls using Azure Purview, dataset versioning and access policies to ensure reproducibility and auditability of ML workflows. Built monitoring and alerting systems using Azure Monitor, improving pipeline stability and enabling proactive issue detection. Conducted A/B testing, uplift modeling and cohort analysis, improving campaign targeting effectiveness by ~23%. Delivered ML-driven forecasting dashboards using Power BI, Tableau and Azure Synapse Analytics surfacing forecast confidence intervals and risk indicators to merchandising and supply-chain teams. Collaborated with merchandising, supply-chain and marketing teams to translate ML forecasts into actionable pricing and inventory decisions, reducing overstock and improving planning accuracy. Implemented version-controlled analytics and ML workflows using Git, Azure DevOps, establishing reproducible model development and deployment standards. Technologies: Python, PySpark, SQL, Azure (Data Factory, Synapse Analytics, Data Lake Storage, Functions, Monitor, Purview, DevOps), Snowflake, dbt, TensorFlow, PyTorch, XGBoost, Random Forest, ARIMA, Prophet, LSTM, Tableau, Power BI, Alteryx, pandas, NumPy, Git, Jenkins ________________________________________ Client: Allstate, Northbrook, IL September 2015 - December 2018 Role: Machine Learning Engineer Developed and deployed supervised machine learning models using Python (scikit-learn, pandas), Random Forest, Gradient Boosting and Logistic Regression for customer churn prediction, fraud detection and insurance risk assessment across policyholder and claims data. Built fraud detection and risk classification models that improved detection accuracy by ~18%, reducing false positives and improving investigative efficiency. Developed scalable feature engineering pipelines using SQL and Python to extract policyholder and claims-level features. Developed actuarial pricing and risk scoring models using GLMs and statistical regression techniques to support underwriting decisions and optimize premium pricing across auto, home and life insurance lines. Built automated ETL pipelines using AWS Glue, Python and SQL to ingest and transform large-scale insurance datasets for model training and batch inference. Trained and validated ensemble models (XGBoost, Gradient Boosting, Random Forest) using cross-validation, hyperparameter tuning and holdout testing to ensure model generalization and production readiness. Implemented comprehensive model evaluation frameworks tracking precision, recall, AUC-ROC and calibration metrics before deployment into underwriting and claims scoring systems. Deployed ML models to production using AWS Lambda, EC2 and batch scoring workflows, enabling automated risk scoring across claims and underwriting operations. Built model monitoring and drift detection scripts using Python and SQL to track prediction distributions and feature stability in production environments. Applied advanced statistical modeling using SAS (PROC LOGISTIC, PROC REG, PROC GENMOD) for actuarial analysis, renewal likelihood modeling and retention forecasting. Optimized large-scale feature extraction queries on Amazon Redshift, reducing data processing time by ~30% for model development workflows. Implemented customer segmentation models using K-means clustering and dimensionality reduction techniques to support retention and cross-sell campaigns. Collaborated with actuarial, underwriting and data engineering teams to translate regulatory and business requirements into deployable ML solutions within a regulated enterprise environment. Technologies: Python, SQL, SAS, scikit-learn, pandas, NumPy, XGBoost, Random Forest, Gradient Boosting, Logistic Regression, K-means, AWS (S3, Redshift, Athena, Glue, Lambda, EC2, CloudWatch)________________________________________ Client: Ooma Inc, Sunnyvale, CA February 2013 - August 2015 Role: Python Developer / Data Analyst Developed and maintained scalable Python-based data processing systems to ingest, transform and standardize telecom usage, call quality and customer interaction data across SQL Server, MySQL and Oracle databases. Developed modular ETL pipelines using Python and SQL to automate recurring data workflows. Built internal analytics services and reporting pipelines supporting churn analysis, service reliability tracking and operational KPIs used by engineering and operations teams. Developed predictive models using Logistic Regression and Decision Trees to forecast customer churn risk and call drop probability, enabling proactive retention and network optimization strategies. Implemented data validation and reconciliation scripts in Python, improving data accuracy by ~30% and ensuring cross-system consistency. Optimized SQL queries and indexing strategies to improve analytical query performance and dashboard responsiveness across large telecom datasets. Built and executed A/B testing frameworks to evaluate feature rollouts and service improvements, supporting data-driven product decision-making. Conducted VoIP and call-quality analytics to identify network bottlenecks and latency issues, providing actionable insights for infrastructure planning. Built executive and operational dashboards using Tableau and Power BI, translating technical metrics into business-impact insights. Collaborated with engineering and operations teams to integrate analytics outputs into production monitoring workflows and customer support processes. Maintained documentation for ETL logic, data models and reporting definitions to support knowledge transfer and long-term system scalability. Technologies: Python, R, SQL (SQL Server, MySQL, Oracle), pandas, NumPy, Logistic Regression, Decision Trees, Tableau, Power BI, ETL Pipelines. PRIORITY PROJECTS LLM Application & RAG Platform Financial Domain Technologies: Amazon Bedrock (Claude 3 / Titan), LangChain, LangGraph, OpenSearch, FAISS, Pinecone, SageMaker, EKS, MLflow, FastAPI Hybrid retrieval architecture combining BM25 and vector search for grounded financial document responses Parameter-efficient fine-tuning using LoRA / QLoRA with PyTorch and Hugging Face Transformers Multi-agent orchestration pipelines for ingestion, retrieval validation and structured synthesis Embedding pipelines with domain-specific chunking and semantic re-ranking Secure FastAPI inference layer with OAuth2 and RBAC controls Containerized LLM workloads deployed on EKS with autoscaling policies LLM evaluation pipelines with automated benchmarking and human validation Observability dashboards tracking latency, token usage, drift and failure rates Clinical GenAI & Risk Modeling Platform Healthcare Technologies: Vertex AI, MedLM (Med-PaLM 2), BigQuery, Cloud Dataflow, GKE, TensorFlow, PyTorch, MLflow Real-time clinical documentation assistant using LLM-based summarization workflows Patient risk prediction models using ensemble learning and deep neural networks Distributed ETL pipelines processing structured and unstructured healthcare datasets Feature engineering workflows integrating social determinant datasets Explainability pipelines using SHAP for model transparency and auditability Containerized inference services deployed on GKE Model lifecycle management with experiment tracking and version control Drift monitoring and performance dashboards using Cloud Monitoring EDUCATIONAL DETAILS Master of Science in Computer Science - University of Central Missouri (Aug 2011 - Jan 2013) Bachelor of Technology in Computer Science - Lovely Professional university (Aug 2007 - Jun 2011) CERTIFICATIONS AWS Certified Machine Learning - Specialty AWS Certified Generative AI Developer - Professional Microsoft Certified: Azure AI Engineer Associate Google Cloud Professional ML Engineer PyTorch Developer Certification Keywords: continuous integration continuous deployment artificial intelligence machine learning business intelligence sthree rlang trade national California Illinois Minnesota New York Tennessee