Mehmet Yilmaz

WORK EXPERIENCE

eBay ― Software Engineer
May 2021 - Dec 2024
Remote

  • Enhanced the Storage Management System (STMS) project to ingest 1.5+ million metrics per minute and growing. Accounting for 70+ arrays, 60+ switches, 1100+ hosts, 900+ disk groups, and 200+ clusters in 3 of eBay’s data centers.
  • Managing and deploying over 45 services across 3 regions on eBay’s Tess.IO cloud platform, ensuring high availability with failover for STMS.
  • Developing and maintaining microservices for processing hardware metrics, facilitating CRUD operations, and providing real time data access to other internal eBay services. Utilizing JavaScript, Node.js, Express.js, Python, Flask, Mongoose, Postgres, and Sequelize to craft efficient RESTful APIs.
  • Spearheaded the migration from MongoDB to Postgres for all STMS microservices, converting 36 collections into 74 tables, integrating Sequelize ORM and custom middleware for a seamless transition.
  • Undertook and improved log monitoring services for Hitachi arrays and Brocade switches, ensuring prompt alerts to the storage team via email, Slack, or PagerDuty through the use of Python and Rsyslog.
  • Leveraging Docker, eBay’s container registry, and Kubernetes to streamline the containerization, deployment, and management of microservices, ensuring the scalability and high availability of STMS.
  • Built a critical backend service for eBay’s Software Defined Storage (SDS) platform, centralizing data for 35K+ volumes, becoming the main data source for volume management, and supporting 780+ platform users.
  • Led cross functional initiatives between STMS, SSI, and EOS teams driving technical improvements and delivering robust, scalable backend solutions to streamline storage operations and data visibility.
  • Developed Restify, an Express.js backend framework, transitioning from Mongoose/MongoDB to Sequelize/Postgres. Inspired by Express-Restify-Mongoose, it specializes in GET requests, offering advanced query features like sorting, selection, limiting, distinction, and inclusion. Additionally, it includes a JSON to Sequelize query translator.
  • Created a findOneAndUpdate() JavaScript function specifically for Postgres, inspired by Mongoose’s equivalent, and using Sequelize. The function was optimized for relational databases with efficient handling of columns, one to one columns, and one to many columns updates, while minimizing database requests by updating only the necessary fields.
  • Built an efficient and accurate native JavaScript package for deep comparison of object properties for both simple and complex nested key value objects.

Hitachi Vantara ― Technical Consultant
May 2021 - Dec 2024
Remote

  • Worked full-time as a contracted software engineer at eBay.

Osgil Defense ― Founder
May 2024 - Oct 2024
Remote

  • Secured $145,000 in pre-seed funding through the Forum Ventures accelerator with a $1.93M valuation.
  • Built and launched TARS, an open-source AI-driven cybersecurity tool for automated threat assessment and penetration testing gaining 275+ stars and 35+ forks on GitHub.
  • Developed PAX, a Python framework for AI-based trustworthiness scoring, reducing hallucinations by leveraging self-reflection, probabilistic prediction, and observed consistency.
  • Led early stage technical development managing teams to reach milestones, build core tech, and establish the company’s initial web presence for outreach and growth.

Anarchy (YC W23) ― Engineering Intern (Part-Time)
Nov 2023 - Jan 2024
Remote

  • Won first place at Anarchy’s October 2023 hackathon with InsightRed, a Reddit marketing tool integrating GPT-4 and Pinecone, leading to my hire.
  • Developed an enhanced API key authentication system for Anarchy’s Chat.dev project, ready for future integration, featuring one-way hashing, support for multiple keys per account, and efficient validation through checksum and caching to minimize database querying.
  • Built llm-speed-benchmark, a benchmarking tool for evaluating the performance of open source LLM models; measuring tokens per second, GPU, CPU, and memory usage, along with runtime.

Colorado School of Mines ― Research Assistant
Feb 2021 - Sep 2021
Golden, CO

  • Undergraduate Research Fellow at the Human-Centered Robotics Lab, mentored by Dr. Hao Zhang.
  • Enhanced the functionality of triangular omni-wheel ground robots, called Tritons, using Python and ROS, enabled precise and efficient movement, accurate rotation, simultaneous control, and a 4x increase in overall movement speed.
  • Implemented proportional feedback systems using Optitrack Cameras and ROS, enabling the Triton robots to precisely reach a specified real-world coordinate with a tolerance of +/- 0.05m within a 2m by 2m area.
  • Successfully transitioned all documentation from Google Drive and Microsoft 365 to GitLab, converting files into Markdown format, and streamlining access and collaboration within the HCR Lab.
  • Mentored a high school student in the fundamentals of programming and robotics, guiding her through the development of a ROS node for controlling the LED colors on the Triton robot.

Namasté Solar ― Intern
Sep 2017 - May 2018
Denver, CO

  • Developed an Excel tool through VBA to analyze company cash flow and employee performance.
  • Worked on designing optimal positions of solar panels on house plans through AutoCAD.

National Oceanic & Atmospheric Administration (NOAA) ― Intern
Jun 2017 - July 2017
Denver, CO

  • Worked with NOAA’s Science on a Sphere (SOSx) team.
  • Developed interactive educational content on the Apollo 11 mission for SOSx using NOAA’s Tour Builder software.

Erols Tailoring ― Customer Service
Sep 2008 - May 2018
Denver, CO

  • Provided exceptional customer service as a part of our family business, greeting, assisting, and collecting payments from customers.
  • Maintained a clean and organized workspace, managing customers’ fitted clothing and ensuring a presentable shop.
  • Responded promptly to customer inquiries and orders over the phone, providing accurate and helpful information.
  • Contributed to the success of the business during my spare time after school, on some weekends, and during some holidays.

Denver Country Club ― Tennis Court Maintenance
Jun 2015 - Aug 2015
Denver, CO

  • Cleaned indoor and outdoor tennis courts which consisted of hard, synthetic, and gravel surfaces.
  • Used the following tools for the job: golf carts, leaf blowers, tennis court squeegees, and gravel rakes.

AWARDS & ACHIEVEMENTS

8090 Top Coder Challenge ― 7th Place
June 2025

  • Placed 7th out of 425 engineers in an 8-hour challenge focused on reverse engineering a black box legacy system.
  • Developed a solution combining classical machine learning techniques with heuristics to replicate a 60-year-old travel reimbursement system.
  • Achieved high accuracy in matching the system’s behavior using only historical data and employee interviews.

National Security Hackathon ― Special Prize Winner
April 2025

  • Won $5,000 Special Prize at Cerebral Valley’s 2nd National Security Hackathon in San Francisco.
  • Led development of Guardian Grid, a smart city emergency evacuation tool for coordinating safe routes during disasters.
  • Successfully completed a special cybersecurity challenge for Code Metal, demonstrating advanced system penetration capabilities.

Anarchy Hackathon ― 1st Place
October 2023

  • Won first place with InsightRed, a Reddit marketing tool integrating GPT-4 and Pinecone.
  • Built a system for identifying potential customers through Reddit comment analysis.
  • Victory led to a part time engineering internship position at Anarchy.

TECHNICAL SKILLS

  • Languages: Python, JavaScript, Bash Script, Rust, Go, Java, C++, & C
  • Databases: Postgres, MySQL, Pgvector, Pinecone, & MongoDB
  • Development: Node.js, Express.js, Flask, Sequelize, Streamlit, Mongoose, Langchain, BeautifulSoup4, Selenium, Axios, yt-dlp, & ffmpeg
  • Cloud: Docker, Kubernetes, Tess.IO (eBay’s cloud platform), Supabase, Vercel, Nginx, AWS EC2, AWS S3, & AWS Lambda
  • Tools: Git, GitHub, GitLab, Cursor IDE, VSCode, Vim, ChatGPT, Claude, Perplexity, V0, Cha, Apollo, Postman, Jira, Slack, Discord, & Ollama
  • API: OpenAI, Anthropic, Cohere, Groq, Travily, Brave, Google Gemini, xAI, & DeepSeek
  • Models: GPT-4o, GPT-4.1, O3, Gemini 2.5 Pro, Claude 3.5 Sonnet, Claude 4 Sonnet, Grok 3, GPT-4.1-Mini, GPT-4o-Mini, DeepSeek R1, Qwen, Llama2, Llama3, & Whisper

EDUCATION

Colorado School of Mines

  • Degree: Bachelor of Science in Computer Science
  • Date: Aug 2018 - May 2022
  • Location: Golden, CO

Denver East High School

  • Degree: High School Diploma
  • Date: Aug 2014 - May 2018
  • Location: Denver, CO

SIDE PROJECTS

Guardian Grid

  • Built a smart city emergency evacuation tool that helps communities respond quickly and safely during major disasters.
  • Developed a system allowing city administrators to mark danger zones and define safe zones on an interactive map.
  • Implemented advanced path finding algorithms to generate and communicate the safest evacuation routes for residents.
  • Created a web portal that uses current location to display danger aware routes to safety with Google Maps integration.
  • Engineered the solution during the 2025 National Security Hackathon, winning a special prize for innovative approach.

PAX (Proactive Agent for eXemplary Trustworthiness)

  • Developed a research agent and framework for measuring and improving LLM response trustworthiness.
  • Implemented systematic approaches for enforced attribution, probabilistic confidence scoring, and observed consistency.
  • Created a self assessment system using multiple LLMs to review and validate responses.
  • Built as a pivot from TARS to address fundamental issues of AI trustworthiness and reliability.

Cha

  • Developed an open-source command-line tool that simplifies interactions with AI models, allowing efficient engagement with powerful language models directly from the terminal.
  • Implemented multi-platform AI support with OpenAI-compatible APIs, enabling users to switch between different AI providers seamlessly.
  • Built comprehensive features including web scraping (YouTube, PDFs, general web content), answer search engine similar to Perplexity AI, interactive code generation with verification, and chat history management.
  • Added advanced capabilities like token estimation, file type support (PDF, DOCX, XLSX, images), integrated shell access, and local configuration options for enhanced workflow efficiency.
  • Recognized with 56+ stars and 4+ forks on GitHub, demonstrating its value to the developer community.

DocuInsight

  • Addressed the “Agreement Trap” by ensuring both signers and senders fully understand contract details to foster transparency and reduce legal disputes.
  • Developed an AI-powered platform that simplifies legal contracts by highlighting key clauses and translating them into plain English, integrating seamlessly with DocuSign for e-signatures.
  • Utilized OpenAI’s cutting-edge models, gpt-4o-mini for efficient JSON extraction and o1-preview for comprehensive document analysis, to generate insights and summaries.
  • Engineered both the backend using Python and Supabase-hosted PostgreSQL, and the frontend with Next.js, deploying on a Linode instance for robust, scalable performance.
  • Integrated Resend’s API for efficient email notifications, enhancing user interaction and contract management workflows.
  • Created for and participated in the DocuSign Hackathon: Unlocked, with a detailed submission on Devpost focusing on unlocking agreement innovation and addressing inefficiencies in contract management through AI.

Notify Cyber

  • Developed and launched Notify Cyber, a platform that aggregates the latest cybersecurity news, achieving 17K+ visitors and 43K+ page views.
  • Generated over 65K views on Reddit with a 96% upvote rate through a zero-cost marketing campaign, growing the waitlist for new features by 160+.
  • Developed and implemented a Postgres database on Supabase, automating content curation with web scraping and summarization using OpenAI’s gpt-4o-mini model.
  • Optimized operations to reduce monthly costs to just $1.85, covering database, API, and hosting expenses.
  • Engineered a Python web scraper to extract and process web articles, leveraging OpenAI’s ChatGPT API for summarization, ensuring efficient database population.
  • Designed, deployed, and managed the project’s Postgres database hosted on the Supabase cloud service.

Threat Assessment & Response System (TARS)

  • Developed an AI-driven tool in Python and Shell script, designed to automate cybersecurity penetration testing.
  • Contributed to securing pre-seed investment for Osgil Defense and amassed +245 stars and +30 forks on GitHub.
  • Utilizes CrewAI, LangChain, OpenAI’s GPT-4 models, Docker, and integrates tools like Nettacker, Rustscan, Zed Attack Proxy (ZAP), and Nmap for comprehensive threat analysis. And applied Streamlit to handle the user interface.

Moving Pose

  • Worked on a three-person class project, implementing the paper “The Moving Pose: An Efficient 3D Kinematics Descriptor for Low-Latency Action Recognition and Detection” by Dr. Zanfir, Dr. Leordeanu, and Dr. Sminchisescu.
  • Set up the Xbox 360 Kinect sensor for real-time skeleton data acquisition using C++ and the Kinect v1.8 SDK.
  • Developed the UI by using Python, Tkinter, C++ and adapting a source code sample from the Kinect V1.8 ToolKit.

SVM On Skeleton Data

  • Developed a college project predicting human behaviors using Python, reformatting MSR Daily Activity 3D dataset to RAD and HJPD forms, and leveraging libraries like pandas, matplotlib, scipy, numpy, and libsvm.
  • Implemented a support vector machine to classify specific human activities/actions, achieving an accuracy range of 62.5% to 70.83%.
  • Demonstrated project execution with successful outcomes, showcasing accurate predictions of human behaviors within the specified accuracy range.

Simple Lane Detection

  • Developed a simple lane detection system using classical computer vision techniques, such as AOI, thresholding, Canny line detection, Hough line transform, and point clustering.
  • Implemented the project in Python, leveraging the numpy and OpenCV libraries to process video data and detect the left and right lanes from the point of view of a driving car.
  • Acknowledged limitations in the method due to video data quality, objects in the area of interest, and road conditions, while gaining valuable learning experience and a solid introduction to computer vision.

Activities

Mines Robotics Club - AgBot Team
Sep 2018 - Jan 2020
Golden, CO

  • Worked on the code for the 2019 AgBot autonomous agricultural robot, which used Python, C++, & ROS Kinetic Kame.
  • Developed a ROS publisher for detecting when the robot has reached the border of a crop field. The code was developed in Python and utilized the robot’s Lidar.