Areg Gevorgyan Resume

Education

University of Maryland, College Park

Expected May 2028

Senior standing.

Relevant coursework:

  • Advanced Algorithms
  • Honors Linear Algebra
  • Real Analysis
  • Probability Theory
  • Mathematical Statistics
  • Programming Languages
  • Abstract Algebra
  • Field Theory
  • Set Theory

Experience

Co-founder at CallCode

January 2026 - present

  • Built parallel AI agent orchestration tool that I am scaling and iterating on; tested with 40 users who built real applications including disability communication and art gallery applications
  • Cursor DC hackathon 1st place, Startup Shell pitch competition winner

Machine Learning Intern at University of Maryland

May 2025 - August 2025

  • Worked on the DARPA Triage Challenge with the Collective Dynamics and Control Lab
  • Built benchmark for testing vision language models and implemented VLM system with 30% higher accuracy and 50% faster on Boston Dynamics Spot robot
  • Utilized Python, HF Transformers, PyTorch, vLLM

Software Engineering Intern at University of Maryland

June 2024 – October 2024

  • Built VR application to educate about the types of lightning at Earth Science Center
  • Presented at AGU 2024 conference and won best presentation award at internship

Fellow at Startup Shell

September 2025 - present

  • Won 6 awards at hackathons and pitch competitions
  • Managing GPU server, NAS, and VPN with 30 users

Projects

MSG: Multimodal Spatial reasoning Gym

2026

  • Built reinforcement learning environment for fine tuning vision language models on 1 million spatial reasoning tasks extracted from hours of construction footage
  • Used to fine tune Qwen 3.5 on 12 H100 GPUs

Cortex

2025

  • Built a machine learning library from scratch utilizing automatic differentiation and implemented neural networks including MLPs and CNNs to successfully train image classifier
  • Taught myself optimization algorithms, backpropagation, neural networks, CNNs, machine learning systems, and machine learning theory

Kubernetes Scaled Inference

2026

  • Built a system to deploy multiple instances of Qwen3.5-0.8B on a server with multiple Nvidia graphics cards
  • Served on OpenAI comparable API to several users

Tomatocode.xyz

2025

  • Built synchronous coding teaching tool that provides live feedback to instructor, currently used in classes

Hicruit.us

2025

  • Built backend for a recruiting tool that automatically screens candidates with AI phone interviews using Python

Linux Shell and Memory Allocator

2025

  • Built a Linux shell and memory allocator from scratch in C utilizing processes, forking, and memory management

Skills

Languages:

Python (advanced), C++, Rust, CUDA, LaTeX (advanced), Lean, OCaml, C, Java, JavaScript, Assembly

Frameworks:

PyTorch, JAX/Flax, Huggingface Transformers, TRL, verl, vLLM, Node.js, Bun

Infrastructure:

Linux, Bash/Zsh, Git, K8s, SLURM, Vim, Emacs, Docker, GCP, distributed multi-GPU training