Avinash Swarna
Dallas,TX
Summary
Machine Learning and AI Engineer with over 6.5 years of expertise in advancing large language model (LLM) architectures and fine-tuning transformer models. Proven track record of enhancing model reasoning through innovative data-driven methods and reinforcement learning techniques. Skilled in developing scalable pipelines and deploying production-grade AI systems that seamlessly integrate cutting-edge research into effective real-world applications. Committed to driving technological advancements that deliver impactful solutions across diverse industries.
Overview
8
8
years of professional experience
1
1
Certification
Work History
Senior ML Engineer
First Citizen's Bank
Dallas, TX
02.2025 - Current
- Developed an enterprise document intelligence platform using Python, LangChain, and OpenAI API to automate contract analysis and legal document processing for Fortune 500 clients
- Implemented advanced RAG architecture using TensorFlow for embedding generation, resulting in improved model performance and more accurate document analysis
- Developed and fine-tuned PyTorch-based LLMs and transformer models for financial use cases such as document intelligence, customer insights, and risk analysis, with a strong focus on accuracy, compliance, and explainability.
- Built scalable GenAI training and inference pipelines in PyTorch, optimizing model performance with mixed precision, GPU acceleration, and efficient fine-tuning (LoRA/PEFT) for enterprise-grade deployment.
- Built vector search functionality using Transformers (Hugging Face) for natural language understanding across 100K+ enterprise documents
- Configured PostgreSQL vector extensions for efficient storage and querying of document embeddings and metadata, leading to faster query response times and improved data management
- Built a production LLM fine-tuning pipeline leveraging Python and Transformers to customize language models for domain-specific legal and financial terminology, integrating data-driven methods to enhance model reasoning and coding capabilities, and improving accuracy by 15%
- Deployed scalable inference infrastructure using FastAPI, Docker, and AWS ECS to serve LLM-powered document analysis with 300ms average response time for enterprise clients
- Established ML model governance framework using MLflow and AWS S3 for model versioning, experiment tracking, and compliance auditing across all LLM applications
- Optimized LLM performance using Redis caching and PostgreSQL indexing strategies to reduce inference latency and improve user experience for document processing workflows
- Collaborated with cross-functional teams on LLM best practices, prompt engineering techniques, and production deployment strategies, leading to improved architectural decisions and successful deployments
- Developed an automated evaluation framework using Python and TensorFlow to assess LLM reasoning quality and output consistency, supporting continuous improvement in model performance
- Implemented enterprise security measures using AWS IAM and encryption protocols to ensure data privacy and compliance for sensitive legal document processing
- Created an LLM monitoring dashboard using Python and MLflow to track token usage, response quality, and system performance across multiple client deployments
- Collaborated with product teams to translate business requirements into technical LLM solutions, focusing on document summarization and clause extraction capabilities
- Established prompt engineering best practices using LangChain and systematic evaluation methods to improve LLM output consistency and reliability for enterprise document analysis
- Built an automated testing pipeline using Python and pytest to validate LLM outputs and ensure consistent performance across different document types
- Contributed to technical reviews and architecture decisions for ML initiatives while participating in technical discussions with engineering teams
Senior Machine Learning Engineer
The Cigna Group
Hyd, India
12.2021 - 08.2023
- Developed a document processing system using Python for automated loan application analysis and data extraction workflows
- Integrated spaCy NLP library for named entity recognition and text parsing from financial documents
- Implemented OpenAI API integration to automate loan application analysis, reducing manual review time by 60% for 5K+ monthly applications
- Built a customer sentiment analysis pipeline using TensorFlow for deep learning model development and training
- Enhanced sentiment classification using Transformers (Hugging Face) to analyze customer feedback and support tickets, achieving 75% accuracy
- Developed automated financial document extraction using Python scripts for parsing bank statements and tax forms, improving data processing efficiency and accuracy
- Implemented spaCy text processing capabilities for employment letter analysis and underwriting workflows, enhancing document analysis speed and accuracy
- Integrated OpenAI API for intelligent document parsing and content extraction from various financial document types, streamlining document processing and reducing manual effort
- Created a customer service chatbot using OpenAI API for natural language understanding and response generation, improving customer interaction and reducing response time
- Built RESTful endpoints using FastAPI to handle routine customer inquiries, resolving 45% of support tickets automatically with a 200ms response time
- Optimized legacy ML models using Scikit-learn and XGBoost for credit risk assessment, improving model performance by 8% while maintaining regulatory compliance requirements
- Configured NLP models using Docker for consistent application packaging and environment management
- Orchestrated scalable document processing using Azure Kubernetes Service (AKS) for handling 20K+ documents daily during peak periods
- Built real-time fraud detection using Python algorithms to process transaction patterns and behavioral analytics, successfully identifying fraudulent activities and reducing false positives
- Implemented a caching layer using Redis for fast transaction lookup and pattern-matching capabilities, significantly improving system response times and user experience
- Configured Azure Database for PostgreSQL for storing transaction data and flagging suspicious activities within 50ms
- Established ML model monitoring using MLflow and Azure Monitor to track model drift, performance degradation, and data quality issues across all production ML systems
- Developed feature engineering pipelines using Pandas and Apache Airflow to process customer behavioral data and alternative credit scoring metrics
- Implemented an A/B testing framework using Python and Scipy to evaluate NLP model performance and measure the business impact of automated underwriting decisions
- Created data validation systems using Great Expectations and Python to ensure financial data quality and regulatory compliance across all ML workflows
- Built API endpoints using FastAPI and Azure Database for PostgreSQL to serve ML predictions to internal applications and third-party partners, improving data accessibility and integration
- Optimized database performance using Azure Database for PostgreSQL indexing and SQL query optimization, enhancing the speed and efficiency of feature extraction from financial transaction histories
- Collaborated with the compliance team to ensure ML models meet regulatory requirements for fair lending and consumer protection in financial services, ensuring compliance and reducing legal risks
Machine Learning Engineer
Value Labs
Hyderabad, India
05.2020 - 12.2021
- Developed customer churn prediction models using Python for data preprocessing and model development workflows, improving customer retention strategies
- Implemented machine learning algorithms using Scikit-learn classification models to identify at-risk customers, enhancing early intervention strategies
- Enhanced model performance using XGBoost ensemble methods, achieving 76% precision in identifying customers 10 days before cancellation
- Built recommendation systems using Python for algorithm development and system architecture design, improving user engagement and satisfaction
- Implemented collaborative filtering using Scikit-learn algorithms to generate personalized product suggestions, increasing clickthrough rates by 12%
- Engineered customer behavior features using Pandas for data manipulation and transformation from e-commerce transaction datasets
- Performed statistical analysis using NumPy for creating 50+ behavioral indicators from purchase history and browsing patterns
- Implemented real-time model serving using Flask APIs for delivering product recommendations to customer-facing applications
- Configured caching layer using Redis to deliver product recommendations with sub-200ms response times for 500K+ daily users
- Automated ML pipeline orchestration using Apache Airflow for scheduling daily model training and feature engineering workflows
- Developed batch prediction systems using Python scripts for processing large-scale customer analytics and scoring, enhancing data-driven decision-making capabilities
- Containerized ML applications using Docker for consistent model deployment across development, staging, and production environments, improving deployment efficiency and reliability
- Built an A/B testing framework using Python and Scipy for statistical analysis to evaluate model performance, leading to data-driven improvements in recommendation algorithms and measurable business impact
- Deployed scalable inference infrastructure on AWS EC2 with auto-scaling capabilities to handle variable traffic loads during the peak shopping period, ensuring seamless user experience and system reliability
- Implemented model monitoring using MLflow for experiment tracking, model versioning, and performance drift detection across all customer-facing ML models, enhancing model reliability and performance over time
- Created customer segmentation models using Scikit-learn clustering algorithms to enable targeted marketing campaigns and personalized shopping experiences
- Established model validation frameworks using Python and Pytest to ensure ML model accuracy and reliability before production deployment
- Optimized feature store architecture using AWS S3 and Parquet format for efficient storage and retrieval of customer behavior features
- Built real-time fraud detection using Scikit-learn classification models and Redis for fast decision-making on payment transactions
- Collaborated with product teams to translate business requirements into ML solutions, focusing on customer lifetime value prediction and inventory optimization
Data Engineer
Infix Healthcare
Hyderabad, India
08.2019 - 04.2020
- Built real-time data ingestion pipelines using Apache Kafka for streaming financial transactions and data from trading systems, improving data processing speed and reliability
- Developed ETL workflows using Python scripting for data transformation and processing automation, reducing manual intervention and increasing efficiency
- Orchestrated data pipelines using Apache Airflow for scheduling and monitoring daily processing workflows, enhancing workflow management and reducing downtime
- Integrated Azure Database for PostgreSQL for customer transaction data storage and regulatory reporting datasets, ensuring secure and compliant data management
- Implemented streaming data processing using Apache Kafka for real-time transaction monitoring and event processing, enhancing the speed and accuracy of transaction analysis
- Built consumer applications using Python for processing streaming data and triggering fraud detection alerts, improving the response time to potential fraud incidents
- Configured a caching layer using Azure Cache for Redis for real-time fraud detection alerts and payment authorization workflows, reducing latency and improving system efficiency
- Built scalable big data processing using PySpark for distributed computing and large-scale dataset analysis, enabling faster data processing and insights generation
- Deployed Spark clusters using Azure HDInsight to handle large-scale financial datasets for compliance reporting and customer analytics, ensuring data accuracy and regulatory compliance
- Optimized database performance using Azure Database for PostgreSQL with advanced indexing strategies and query optimization
- Implemented SQL query optimization techniques for sub-second response times in high-frequency trading environments
- Deployed containerized ETL applications using Docker for consistent application packaging and environment management
- Orchestrated container deployment using Azure Container Instances for consistent job execution across development, staging, and production environments
- Created comprehensive data validation frameworks using Python and the Great Expectations library to ensure financial data accuracy and completeness across all pipeline stages
- Established an intelligent caching architecture using Azure Cache for Redis for frequently accessed customer and transaction data, reducing database load by 40% during peak trading hours
- Automated infrastructure provisioning using Azure Resource Manager templates and Terraform for managing Azure Blob Storage data lakes and compute resources
- Built monitoring and alerting systems using Azure Monitor and Application Insights to track Apache Airflow pipeline health, Kafka processing latency, and system performance metrics
Junior Data Engineer
Infix Healthcare
India
05.2018 - 06.2019
- Wrote complex SQL queries using PostgreSQL for extracting patient data from multiple EMR systems and clinical databases
- Optimized database queries using SQL Server to support clinical reporting and population health analytics for 500K+ patient records
- Developed ETL scripts using Python programming for data extraction and transformation workflows from healthcare systems
- Implemented data cleaning processes using Pandas library to validate patient data from Epic and Cerner systems for downstream analytics
- Automated data validation processes using Python scripts for identifying missing values and duplicate records in clinical datasets
- Performed statistical analysis using NumPy for detecting data inconsistencies and quality issues in healthcare databases
- Created healthcare dashboards using Tableau for visualizing patient flow metrics and resource allocation reporting
- Built reporting queries using SQL to support operations teams with bed utilization and clinical performance metrics
- Implemented healthcare data standards using Python libraries for HL7 message parsing and healthcare data integration projects
- Validated FHIR resources using custom Python scripts to ensure patient data integration compliance and accuracy
- Built HIPAA-compliant data processes using PostgreSQL encryption features for protecting sensitive patient information in development environments
- Implemented data masking using Python scripts and access controls for secure handling of clinical data during testing phases
- Developed Excel-based reporting tools using VBA macros and SQL connections to generate weekly clinical quality reports for hospital administration teams
- Monitored ETL pipeline performance using SQL queries and Python programming frameworks to identify data processing errors and system performance bottlenecks
- Collaborated with clinical analysts using Jupyter Notebooks and Python for ad-hoc data analysis to understand healthcare requirements and translate them into technical specifications
- Maintained comprehensive documentation using Git version control for ETL processes, database schemas, and healthcare data integration workflows
- Performed database optimization using PostgreSQL indexing strategies and SQL query performance tuning under the guidance of senior database administrators
- Built data quality frameworks using Python and basic validation scripts to ensure healthcare data accuracy and completeness across all pipeline stages
Education
Master of Science - Computer Science(Business Analytics)
Trine University
Angola, Indiana, IN12.2024
Skills
- Programming Languages: Python, SQL, Java
- Machine Learning/AI Frameworks: Scikit-learn, XGBoost, TensorFlow,Pytorch, MLflow, RAG, MCP
- NLP & LLM Technologies: LangChain, OpenAI API, Hugging Face Transformers, spaCy
- Vector Databases & Search: Pinecone, Lang Smith
- Workflow Orchestration: Apache Airflow
- Stream Processing: Kafka, PySpark
- Data Processing Libraries: Pandas, NumPy, Scipy
- Data Quality & Testing: Great Expectations, pytest
- Databases: PostgreSQL, SQL Server, Redis
- Cloud Platforms - AWS: EC2, S3, ECS, IAM
- Cloud Platforms - Azure: AKS, HDInsight, Database for PostgreSQL, Blob Storage, Container Instances, Monitor, Application Insights, Active Directory, Key Vault, Purview, Cache for Redis, Resource Manager
- Web Frameworks & APIs: FastAPI, Flask
- DevOps & CI/CD: Docker, GitHub Actions, Terraform, ARM Templates
- Data Visualization & Analytics: Tableau, Jupyter Notebooks
- Office & Productivity: Excel, VBA
- Healthcare Standards: HL7, FHIR, HIPAA, Epic, Cerner
- Documentation & Collaboration: Git
- AI Research & Techniques: Machine Learning, Fine-tuning Language Models, Reinforcement Learning, Developing Benchmarks, Model Optimization, Data-Driven Methods
Certification
AWS certified solutions architect
Timeline
Senior ML Engineer
First Citizen's Bank
02.2025 - Current
Senior Machine Learning Engineer
The Cigna Group
12.2021 - 08.2023
Machine Learning Engineer
Value Labs
05.2020 - 12.2021
Data Engineer
Infix Healthcare
08.2019 - 04.2020
Junior Data Engineer
Infix Healthcare
05.2018 - 06.2019
Master of Science - Computer Science(Business Analytics)
Trine University