Traba is building a world where the global supply chain operates at peak efficiency. We connect industrial businesses with reliable workers to fill open shifts and empower both businesses and workers to reach their full productivity and potential.

Backed by Founders Fund, Khosla Ventures, and General Catalyst.

• Designed and implemented a project that monitors unexpected recovery actions of RDS Aurora instances during workflows, improving visibility on failure trends across millions of instances

• Authored the design document considering alternatives and trade-offs, proposing implementation, testing and deployment strategies, resulting in swift approval from key stakeholders

• Developed the project using Java and Spring with design patterns such abstract factories, composites, adapters, and the strategy pattern to contribute robust and maintainable code

• Tested my code with JUnit and the Mockito framework, reaching upwards of 95% line and branch coverage

• Created a software system that tracked the cost and utilization of machine learning products across 10+ Kubernetes clusters, exposing efficiency metrics for AWS expenditures of more than $1M per year

• Extracted cost and utilization metrics from Prometheus and Yotascale to expose operational insights ranging from multi-cluster aggregates down to individual pods, delivered through SuperSet Dashboards

• Pioneered a novel deployment strategy for tree and forest models by converting them with RAPIDS FIL and deploying them onto NVIDIA Triton Inference Servers, resulting in a 43% inference speed increase for the production payment fraud detection LightGBM model

• Developed modules of the new model evaluation pipeline that allowed for the simultaneous analysis of

various block models, providing greater flexibility to find the most optimized model for desired metrics

• Created the data and model loading module that relied upon regex matching and multiprocessing pool to accurately and efficiently load desired objects

• Leveraged the Python Hashlib blake2b hashing function to implement the result saver module which

identified and cached previously computed metrics, greatly decreasing redundant computations

• Devised a tool that facilitated and improved the query of GitHub PR by caching data in a PostgreSQL database, alleviating traffic on the enterprise servers, and improving the workflow of over 3000 developers at Cisco

• Implemented inbound and outbound backend API using Golang which transformed raw webhook data into categorized table entries and served query results to front end clients

• Designed the frontend experience with Python, Flask, and Bootstrap 4, creating a clean and efficient interface that offered more flexibility and filtering granularity than the native GitHub PR searching tool

• Created a standalone python script that could quickly retrieve and aggregate the mailer of various Github Code Owner groups into an easily consumable JSON, giving developers at Cisco an effortless method of finding and contacting the owners of any file in their extensive codebase