Davit Barblishvili Email & Phone Number

New York, United States

New York, NY, US

• Team: ML Observability
• Developing a service to evaluate LLM applications by detecting key metrics such as toxicity, failure toanswer, language mismatches, and prompt injection, providing users with out-of-the-box insights into.
• Engineering a configurable model-based evaluation platform that allows users to select LLM providers,models, and custom prompts, delivering tailored insights into the performance of LLM applications.
• Building a Golang-based LLM Observability Event Processor that consumes Kafka messages, computesembeddings using OpenAI and in-house models, and analyses records for advanced data analytics.
• Designing and implementing a custom rate limiter in Golang, leveraging Redis for distributed statemanagement to enforce unique rate limits across multiple services, optimising resource utilisation andenhancing system.
• Conducting sentiment analysis on user-chatbot interactions using both open-source and LLM-basedmodels, achieving an F1 score of 70%.

New York, NY, US

• Team: Trust & Safety
• Led the development of the Datadog Safety Center using React and Golang, creating a comprehensivesolution for end-to-end assessment, visualisation, and management of critical security configurationswithin Datadog.
• Designed and implemented a Java-based reducer that analyses audit logs to swiftly identify and alert onabnormal activities, such as new device or new country logins enhancing user account security.
• Developed a Python T oken Leak Detection service for continuous monitoring of secrets, incorporatingautomated breach notifications and token revocation, reducing revocation time to 2 days.
• Created a Golang-based Enrichment Service serving as a gRPC API to aggregate Threat Intelligencedata, monitoring for malicious endpoints, IP addresses, and compromised passwords.
• Engineered email verification REST API endpoints in Python and React, seamlessly integrating with th

Austin, Texas, US

• Developed a cross-platform mobile application for Android and iOs using Flutter Development Kit.
• Utilized Bloc architecture for the state management to successfully connect presentation layer and data layer.
• Rendered Auto-CAD 3D objects using Pytorch3D and Plotly libraries.
• Built an object detection algorithm using convolutional neural networks (VGG16, ResNet) to detect,recognize, and identify entities in the image.
• Constructed scalable and secure APIs using gRPC, Protocol Buffers, and Golang.

New Brunswick, NJ, US

• Researched object detection/classification technologies to implement human detection models to analyze video frames optimizing prediction of human traffic patterns by 10%.
• Engineered asynchronous processing of image frames from video source to Azure storage using multithreading and Asyncio module decreasing the frame processing/upload time by 1/2.
• Utilized Databricks to process and transform massive quantities of image frames to parallelize and improve existing computer vision models to efficiently predict human presence in image frames.
• Deployed the model on Raspberry Pi and Nvidia Jetson TX2 to reduce the operational cost by 50%.

San Francisco, California, US

• Developed/Maintained client-side single page web applications using ReactJS and Redux libraries.
• Utilized various technologies to perform code repairs and optimization of web applications for customers.
• Created and executed automated software test plans, cases and scripts to uncover, identify and document software problems and their causes.

New York, New York, US

• Built an application that enables doctors to find accurate information on FDA approved drug treatments.
• Utilized Scrapy, Splash, and Selenium to extract desired data from a given URL in an automated way and using BeautifulSoup to parse and clean XML/HTML files reducing human engagement by 50% in search tasks.
• Structured back-end NoSQL database in Google Firestore for complex and optimized queries.

New York, NY, US

• Assisted Professor of Discrete Mathematics to supervise a class achieving 20% increase in the student’s engagement.
• Taught logic, formal proofs, graph theory, number theory, probability and statistics to the students during office hours increasing average class performance by 1/2 letter grade.
• Guided students on a weekly base to complete programming projects in Python generating 20% increase in class performance compared to previous semesters.

New York, NY, US

• Utilized Excel to manipulate large amounts of data in order to perform data analysis and data manipulation to produce complex reports and efficient dashboards.
• Automated cleaning of unstructured data using Power query and Advanced Editor generating organized data structure.
• Integrated/merged multiple financial reports into a single report using Power Query and Pivot Table.