Charlotte, North Carolina, United States
šØ Currently doing: I am a Sr ML Engineer with Judi Health (CapitalRx) working on conversational voice AI. š” Background: Extensive experience with AI/ML Engineering, API Development, Cloud Computing, Big Data My background spans many domains- AI/ML Engineering (AWS SageMaker, AWS Bedrock, GCP Vertex AI, Huggingface), API Development (FastAPI), AI/ML Platform (MLFlow), Cloud Computing (IAM Persona Management, Terraform Modules, Service Catalog Products), CICD (AWS CodePipeline/CodeBuild, Jenkins) and Big Data (EMR, Pyspark, Airflow) - all geared towards the data science domain. š In my free time you can catch me listening to Lo-Fi and cooking carbonara
ā¢ Lead two engineers to develop a full-stack Agentic AI application, architecting a document processing pipeline using Docling, Streamlit for frontend, vector search using FAISS, Llama Index for AI, and AWS Bedrock for inference. ā¢ Deployed a RAG API for Underwriter Guideline Doc Q/A, with optimized GPU usage for Flan T5 XXL using Sagemaker Deep Java Learning Containers, which reduced latency by 10x relative to previous solution using GPT 3. Success led to organization-wide push to adopt and build GenAI solutions. ā¢ Architect and develop a production-ready AI/ML Application Platform on AWS featuring containerization, load balancing, SSO Integration, and high availability. The platform supports 30+ applications in production across multiple business verticals. ā¢ Lead AI Platform eng for first-in-company Google ADK Agent API with a Session persistence layer, enabling scalable end-to-end system functionality using Terraform, CloudRun, FastAPI, AlloyDB, and a Load Balancer. ā¢ Upgrade CICD pipeline product to support pyproject.toml python builds Expanding on foundational AI Platform via Agentic AI.
ā¢ Served as Lead Engineer for AWS Bedrock-Splunk Observability integration, implementing a scalable event-driven solution using SQS, Lambda, and S3. Used by all enterprise AI workloads while adhering to strict enterprise data-boundary standards. ā¢ Lead AI customer enablement initiatives as embedded technical consultant across 10+ projects, evangelizing and implementing AWS, Arize, MLFlow, CICD (CodePipeline/Jenkins), unit testing, and MLOps best practices to dramatically enhance data science lifecycle processes. ā¢ Built a successful POC of LLM Insurance Document Processing leveraging Chain-of-Thought Prompting, Pydantic, and Anthropic Claude 3, resulting in full-scale project approval and dedicated resource allocation. ā¢ Fine-tuned DinoV2 Vision Transformer for fraudulent image classification hackathon using Huggingface; won 3rd place in org-wide hackathon. Consolidating and hardening MLOps Platform.
ā¢ Built and deployed production ready industry classification NLP model, leveraging Apache Airflow for orchestration, and SageMaker Batch Transform for optimized inference across large datasets ā¢ Developed a Transient EMR Service Catalog product to enable federated, self-serve distributed Big ā¢ Data Processing. Integrate EMR with various data stores like Snowflake, S3, Hive, and Oracle. ā¢ Developed high-availability Apache Thrift servers to enable secure access to Hive Databases via Transient EMR. ā¢ Project lead in refactored data science IAM Console Access Controls to align with SDLC for data science use cases. Utilized Boto3 to migrate over 1000+ service catalog products with no impact to end users. Zero to One buildout of MLOps Platform on AWS
Directly collaborated with insurance-tech startup founders for go to market in the United States. ā¢ Worked closely with founders to act as Voice of Customer to refine product experience. Conducted customer interviews and identified key product pain points ā¢ End to end management of marketing campaign execution and strategy for GTM. Increased Instagram following by 200% and lead to net increase in sales ā¢ Performance market research to discover lucrative market niche to pursue. Lead to increased app download rate ā¢ Vendor management and coordination with international internal teams to ensure smooth alignment on marketing
Automate SOX Database reporting from ~60 days to 30 minutes. ā¢ ETL of Oracle database roles & automated the identification of highly privileged database roles for compliance and reporting with SOX . Improved on report generation time by ~100 days. Tech Stack: Pandas, CX_Oracle, Oracle Enterprise Manager, Toad Data Point, Oracle SQL
Adhoc data science projects (survival analysis, customer churn analysis, etc) ā¢ Conducted survival analysis using the Lifelines library to quantify and predict customer churn rate ā¢ Executed the end to end analysis of vendors product recommendation algorithm effect on ecommerce ā¢ Automated the analysis of website text fields to find typos & recommend them for removal ā¢ Visualized data + communicated project conclusions through creating python plotly-dash dashboard ā¢ Examined program/application code (C#/VB) to create a data dictionary of identified SQL servers, tables, and stored procedures to aid IT team in cloud migration Tech Stack: Requests, Pandas, Matplotlib, Plotly, Seaborn, Dash, MS-SQL/SSMS, Lifelines