Projects
Selected work across generative AI, contact center AI, machine learning, and cloud architecture. Client names are withheld.
Agentic AI workflows
Autonomous, multi-step AI agents that reason over enterprise knowledge and act through existing systems, with retrieval grounded in Amazon Bedrock Knowledge Bases.
Stack: Amazon Bedrock, Amazon SageMaker AI, LangChain/LangGraph, Bedrock Knowledge Bases
Generative AI voice assistant
A contact center voice bot that answers caller questions from company documents using retrieval-augmented generation.
Stack: Amazon Connect, Amazon Lex, AWS Lambda (Python), Amazon Bedrock, Amazon OpenSearch Service, Amazon S3
Contact center applications and survey workflows
Post-call customer survey flows and contact center workflows, migrated from a third-party platform to AWS-native services.
Stack: Amazon Connect, Amazon Lex, AWS Lambda (Go), Amazon Kinesis, AWS Step Functions, Amazon DynamoDB, Amazon S3, Terraform
Fraud detection model
A machine learning model that flags potentially fraudulent property claims, taken from feature engineering through training, tuning, and deployment.
Stack: Amazon SageMaker AI (XGBoost), Amazon S3, Jupyter
Fine-tuned language model
RoBERTa-large fine-tuned on domain-specific data for a specialized NLP task.
Stack: Amazon SageMaker AI JumpStart, Hugging Face, Amazon Bedrock
Private RAG system for sensitive data
A retrieval-augmented question-answering system over medical documents that runs entirely on local infrastructure, so no data leaves the environment.
Stack: Python, gpt-oss-20b, LM Studio, ChromaDB, LangChain
Serverless data and API platforms
Event-driven APIs and streaming pipelines with search and analytics, fully defined as infrastructure as code.
Stack: Amazon API Gateway, AWS Lambda, Amazon DynamoDB, Amazon S3, Amazon Kinesis Data Streams, Amazon Data Firehose, Amazon OpenSearch Service, Terraform, AWS CDK