01.

About

B.S. in Computer Science & Mathematics from Whitworth University.

Focused on backend development, cloud infrastructure, and security engineering. Building systems with dependable infrastructure and zero-trust architecture.

bereket.ts
📍 Location
>Software Engineer based in Seattle, WA.
🤖 Currently
>Competing in the OpenAI Parameter Golf Challenge — training a language
>model to fit within 16MB, under 10 min on 8x H100 GPUs, optimized for
>compression on FineWeb.
🏆 Achievements
>Celeri.io — won $50K at Sparks Weekend to build communication software
>for criminal courts, reducing pretrial detention times.
>3rd Place — ICPC Pacific Northwest Regional 2025.
💼 Experience
>Security Engineer Intern @ Washington Trust Bank.
>Full-Stack SWE Intern @ Hewitt Learning.
02.

Experience

Mar 2026 — PresentRemote

Parameter Golf Competitor

OpenAI

Training a language model to fit within 16MB, under 10 minutes on 8 H100 GPUs, optimized for maximum compression on FineWeb. Engineering custom training loops with gradient accumulation, mixed-precision arithmetic, and memory-efficient data loading.

PythonPyTorchLLMQuantizationH100
Jan 2024 — PresentSpokane, WA

Startup Engineer (Founding Team)

Celeri.io

Won $50,000 investment at Sparks Weekend to build communication software for criminal courts, reducing pretrial detention times by connecting stakeholders across counties.

Watch the pitch →
Pitch DeckPrototypingTeam communicationProduct DesignLegal-Tech
May 2024 — Aug 2024Spokane, WA

Security Engineer Intern

Washington Trust Bank

Automated IT and vulnerability monitoring workflows using PowerShell and Python REST APIs, eliminating 80+ engineering hours/week. Deployed 5+ automated security compliance checks integrated with Azure and GitHub Actions CI/CD.

PythonPowerShellAzureGitHub ActionsSecurity
Feb 2025 — May 2025Remote

Full-Stack Software Engineer Intern

Hewitt Learning

Designed scalable REST APIs with C#, ASP.NET Core, and MySQL with RBAC. Automated PDF report generation reducing turnaround by 60%. Deployed via GitHub Actions CI/CD.

C#ASP.NET CoreMySQLREST APIsCI/CD
03.

Skills

< />

Languages

8 tools
C++CPythonGoTypeScriptJavaScriptSQLBash
{ }

Frameworks

6 tools
ReactNext.jsFastAPINode.jsASP.NET CoreTailwind CSS

Cloud

9 tools
GCPBigQueryPub/SubCloud RunDataflowVertex AIAWSAzureFirebase
🔒

Security

6 tools
Zero TrustRBACVulnerability AssessmentSIEMPowerShell AutomationInfoSec

Systems

7 tools
LinuxMultithreadingLock-Free DSTCP/UDP SocketsMemory ManagementDockerKubernetes

Data

7 tools
TensorFlowScikit-learnPandasNumPyRedisPostgreSQLMySQL
~/learning
$ always learning
> currently exploring quantitative finance
> systems programming & low-level optimization
> NeetCode 150 DSA — graphs, intervals, daily
$ status: grinding ▮
04.

Beyond the Code

Fitness & Lifting

Tracking every session on Hevy. Consistency over intensity.

Follow me on Hevy →

Soccer

Love playing — I'm down anytime, anywhere. On weekends you'll find me watching the English Premier League.

Anime

Attack on Titan changed the way I think about freedom, sacrifice, and what it means to fight for something bigger than yourself.

Video Games

Far Cry 3 for the storytelling — Vaas is one of the greatest villains ever written. Uncharted 4 for the world building and cinematic feel.

Spiritual Life

Actively working on my spiritual life. Faith keeps me grounded.

Coffee

You can't represent Ethiopia without loving coffee. I'm all about that sip-and-code life.

05.

Projects

DevScope — Repository Intelligence Platform

Built a distributed engineering analytics platform mining GitHub repository data to surface PR velocity, review latency, and code churn. Streams events through Cloud Pub/Sub into BigQuery for sub-second querying. Vertex AI anomaly detection on Cloud Run flags productivity regressions with 95% precision.

PythonGCPBigQueryPub/SubVertex AICloud RunReactTypeScript

Network Intrusion Detection System

Real-time network intrusion detection system classifying live traffic across 7 attack categories including DoS, DDoS, port scanning, and brute-force. Trained Random Forest on 2.5M+ CICIDS2017 records achieving 97.47% accuracy and 98.21% F1-score. Results visualized in a live Streamlit dashboard.

PythonScikit-learnRandom ForestStreamlitCICIDS2017

CPU Scheduler Simulator

Implemented FCFS, Round Robin, and MLFQ scheduling from scratch in C and benchmarked against 1,000+ real PlanetLab distributed VM workload traces. Measured throughput, CPU utilization, turnaround time, and response time to identify optimal strategy per workload type.

CPlanetLab TracesSystemsOS Concepts