Ericson Davis Email & Phone Number

Tysons, Va, Us

Developing Retrieval Augmented Generation (RAG) applications. Researching and prototyping LLM-based orchestration frameworks.

Las Vegas, Nevada, Us

Developing ML and statistical models to drive product search results and customer personalization.Developed embedding-based approach to search query target classification and created the CI/CD pipeline/endpoint using Sagemaker, Lambda, and API Gateway from CDK.Developed new engagement probability-based autocomplete model for Zappos search.Ported existing models to Spark/Glue to increase scalability and reduce runtime.Developed prototype customer embedding model to improve personalized product recommendations.

Seattle, Wa, Us

Helping to optimize/simulate middle mile truck schedules and loads for Amazon's North American supply chain.Developed and deployed a mixed-integer programming approach to shift truck departure times to reduce truck arrival (yard size) and total package receipt constraint violations for prime week 2021. Extended the model to alleviate downstream labor capacity/available volume mismatches for peak 2021.Developed an ECS-backed self-service simulation tool for customers (other middle mile planners) to assess the supply-chain impact of potential middle mile truck schedule changes.Developed and deployed proactive accuracy tracking tools for prediction models using CloudFormation (from jinja)/Glue/Athena/Tableau

Seattle, Wa, Us

Designed and prototyped anomaly detection algorithms in scala/spark to perform daily analysis of a data set exceeding 250 TB.Analyzed all internal AWS fleet utilization statistics using Athena (SQL), spark, and python. Designed and deployed algorithms to generate efficiency recommendations for fleet managers using AWS Cloudformation, Glue, and Cloudwatch.Deployed a prototype machine learning model through an AWS Sagemaker endpoint to identify and classify computer programming languages and structured data formats within free text. The model achieved 98% accuracy during testing.Analyzed customer usage patterns to improve service revenue forecasting.Designed a new probabilistic rack power consumption model to capture all possible power draw scenarios and implemented as Glue(spark)+Lambda pipeline using AWS CDK.Advised ML/Optimization projects to:* Select which AWS services should launch in new region builds* Determine when to announce new feature/service launch dates

Tysons, Va, Us

Collected open source text data through web scraping and other available APIs to support internal human resources related natural language processing projects applying topical clustering, document classification, sentiment analysis, and word embeddings.Developed AWS pipelines (using Lambda, RDS, and EC2) to ETL incoming datasets and process transformed data with existing ML models for visualization in .NET/Tableau server front-end.Automated the production of PNG stock levels for ~500K items for the Defense Logistics Agency reducing processing time from ~2 weeks to 30 minutes. New stock levels contributed to ~$400M in savings.

Arlington, Va, Us

Automated the extraction of contracts-related data by fusing disparate references to data elements from semi-formatted text resulting in a 95% decrease in processing time with 97% accuracy on key data fields.Designed and implemented interactive data visualization dashboards in MS Access and MS Excel to show project progress and legacy weapon system costs.Designed and built a generalized cost model framework in python for use in ERDC's Tradebuilder (weapon system capability/design/cost tradespace analysis).

Reston, Virginia, Us

Developed models/prototypes for release/referral decision classification and automated document redaction.Supported the EMBERS (Early Model Based Event Recognition using Surrogates) project as a developer maintaining open source data feeds on a Linux-based AWS cluster.

M.S., Mathematics

B.S., Mathematics