NAME

Ari (Hamid) Zare - ML and distributed systems

SYNOPSIS

Ari [--engineer] [--developer] [--diy-hacker] [--other]

DESCRIPTION

Staff Software Engineer at Axon AI with a master's degree in computer science from Vanderbilt University. Specialized in MLOps, distributed systems, and machine learning infrastructure. Proven track record of leading technical teams and delivering scalable AI solutions.
Feel free to contact me if you'd like to connect. Email: az@arizare.com

Disclaimer: Due to budget cuts this man page might be out of date. Please refer to the PDF version here.

EXAMPLES

ari --work-history

jump to recent roles and impact

ari --capabilities

scan domains, tooling, and strengths

ari --projects

inspect side projects and hands-on work

ari --contact

open author block and external links

ari --pdf

fetch printable version

INDEX

NAME | SYNOPSIS | DESCRIPTION | EXAMPLES

EXPERIENCE | CAPABILITIES | PROJECTS | EDUCATION

EXPERIENCE

Staff Software Engineer at Axon AI Oct 2025 - Present

Joined via Prepared acquisition in 2025

Staff Software Engineer at Prepared Mar 2025 - Sep 2025

Accelerating emergency response with Assistive AI

Working to ensure emergencies get the best possible response.

Acquired in October 2025

Senior Software Engineer at HPE AI via Determined AI Acquisition June 2021 - Dec 2024

I contributed in various capacities most recently in the resource management team, and our GenAI product. I also took on roles to design and collaborate in launching three greenfield projects in new deep learning SaaS and Generative AI.

Leadership and Mentorship: I introduced and revised engineering processes in various teams. I mentored 10 new hires, conducted over 70 interviews, contributed to customer and product meetings, and consistently put team success over individual recognition.

Some projects:

Led a project to support multi-cluster, multi-resource manager Kubernetes using gateway proxies for Determined tasks, which also led to the development of a remote K8s development tool.
Architected and implemented enhancements to our job scheduler, adding visibility and control which exposed its decision-making process and uncovered obscure bugs, resulting in higher GPU utilization and a 40% reduction in job churn rate.
Overhauled our metrics capturing and reporting system for increased performance and flexibility, utilizing database technologies such as PostgreSQL partitioned tables. This necessitated large-scale migration, performance testing, and optimization.
Led platform integration with the PyTorch Lightning, resulting in 40% fewer lines of code (LOC), fewer user errors, and faster porting time for those models.
Improved fault tolerance in HPC environments with faulty GPUs and hardware by implementing node and slot taints, enabling faster job recovery.
Reduced CLI startup time by 80% through dependency tree analysis and lazy imports.

API: Designed and delivered various public and internal APIs based on standards such as OpenAPI, protobuf, gRPC, and WebSockets. Advocated for and developed various in-house code generators, which enabled many engineering workflows.

Tooling, CI/CD, and Testing: Identified and eliminated technical and process inefficiencies, achieving multiple 10x improvements in PR turnaround time. Initiated and delivered testing and CI infrastructure for various teams.

Software Engineer at Determined AI Aug 2019 - Jun 2021

Developing an end-to-end deep learning pipeline to enable automation of model selection and hyperparameter tuning. Distributed Deep Learning; Horovod; PyTorch; PyTorch Lightning; experiment tracking; hyperparameter search; Domain-aware scheduling; REST, gRPC, and Python API design

Built an open-source end-to-end deep learning pipeline to enable automation of model selection, tracking, and hyperparameter tuning that was acquired by HPE high-performance computing unit in 2021. Read more above.
Joined as engineer #14 and helped build the team and wore many hats based on changing needs of the early-stage startup.

News: Acquired by HPE in 2021 . Link to the press release by HPE

Software Engineer (Intern) at Real-Time Innovations (RTI) Summer 2019

Research and development as part of the core team, working on two projects for the next release of Connext DDS. C, C++, Java.
Designed and implemented APIs to expose information about the type system for customers. C, C++
Evaluate, research, and develop a pluggable lossless compression solution for real-time, peer to peer communication of user data. Targeting use-cases for autonomous vehicles: sensor sweeps and camera feeds.

Software Engineer (RA) at Institute for Software Integrated Systems 2017 - 2019

Developing our open source, web-based visual programming environment, NetsBlox introducing distributed programming concepts like message passing and remote procedure calls to block based languages.

Standardized a base class for developing RPCs, improving performance and reducing lines of code by up to 70%
Performance optimizations: Improved messaging throughput by up to 24x under load. Selective caching, through a layer positioned close to the networking module, in conjunction with smart cache busting enabled the router to avoid stale routing information and provide verifiable socket information for clients at a much higher rate than before.
Developed a parser for dynamic creation of documentation for service APIs currently covering 130 modules and counting
Redesigned user-facing services and RPCs to guarantee type safety in JavaScript removing the burden from the developers.
Reduced production client bugs by 20% through automated browser tests, a chaos monkey and Travis CI.

[Part-time research assistant]

Teaching Assistant at Vanderbilt University Spring 2017

Assisted with Program Design and Data Structures course taught in C++. Helped students understand different data structures and algorithms, design modular programs, use best practices, edge case testing and safety checks.

CAPABILITIES

domains

toolchain

working-style

If it ain't broke, I'll fix it. See the projects section.

EDUCATION

Doctor of Philosophy (Ph.D.) in Computer Science at Vanderbilt University

Advisor: Prof. Ákos Lédeczi.

I decided to drop out of the Ph.D. program end of 2020 to pursue a career in industry.

Master of Science (M.Sc.) in Computer Science at Vanderbilt University

Advisor: Prof. Ákos Lédeczi.

Researching the intersection of Computer Science education, Networks, Distributed Systems, and Robotics in Vanderbilt's cyber physical systems lab, Institute for Software Integrated Systems.

Coursework: Machine Learning; Modeling & Simulation; Deep Learning; Advanced Algorithms; Distributed Systems; Web-based System Architecture; Concurrent Java Programming; Reinforcement Learning; Cloud Computing

Bachelor of Science (B.Sc.) in IT Engineering at Azad University

Coursework included: Data Structures; Software Engineering; Databases; Artificial Intelligence; Computer Networks; Operating Systems; Programming Languages; Statistics & Probability; Operations Research;
* Top Student: Engineering department, junior and senior years.

PROJECTS

Selected work. Details and more in projects(7); older projects remain in the archive below.

Determined - Open-source deep learning training platform for GPU clusters, acquired by HPE (2019 - 2024)

Open-source platform for distributed training, hyperparameter search, and GPU cluster resource management, with a Go master, a Python training harness, and Kubernetes and Slurm resource managers. Joined Determined AI as engineer #14 in 2019 and stayed through the 2021 HPE acquisition, working across the job scheduler, metrics storage, Kubernetes support, APIs, and CLI. Landed 695 merged PRs in the public repo, which has 3k+ GitHub stars. Go, Python, PostgreSQL, Kubernetes, gRPC, TypeScript
github

Multi-cluster Kubernetes gateway proxies - One Determined master running tasks across multiple Kubernetes clusters through gateway proxies (2024)

Feature in the open-source Determined platform that lets a single master schedule tasks across multiple Kubernetes clusters and resource managers. Gateway resources deploy alongside each task's rank 0 pod, and task ports are translated to a configurable range of gateway listener ports so proxied traffic reaches tasks in remote clusters. Led the project at HPE through implementation, tests, and user docs, and spun out a developer tool that connects a local dev cluster to remote Kubernetes clusters. Go, Kubernetes, Helm, Python

mao - Multi-agent orchestrator that fans one prompt across coding agents and reconciles the results (2026)

Python orchestrator that runs CLI coding agents concurrently in fan-out, fan-in, and organizer pipelines, and tracks their sessions across machines. A hosted console authenticates requests and enqueues typed commands into a Postgres registry; per-machine workers claim them with leases, run local agents, and report topology. The browser never executes shell commands or sees secrets, and API contracts are OpenAPI snapshots with generated clients and drift checks. Python, PostgreSQL, FastAPI, Go, React, Docker
live

VirtualCam - Face tracking and background effects for any video call app (2020 - 2026)

Real-time virtual camera pipeline that captures a camera or screen, runs segmentation and face tracking, and outputs through v4l2loopback on Linux or OBS Virtual Camera on macOS. Started in 2020 because Zoom had no virtual backgrounds on Linux; now adds background blur and swap, desk view, gesture reactions, and a native SwiftUI menu bar app that drives a managed Python worker over a JSONL protocol. 71 GitHub stars. Python, OpenCV, MediaPipe, Swift, SwiftUI, v4l2loopback
github | product page | demo video

GenAI Studio - Generative AI playground built on the Determined training platform at HPE (2023 - 2024)

HPE's generative AI playground for deploying and prompting open models from a web UI, built on the Determined deep learning platform. As one of the early senior engineers, contributed to its design, iteration, and productionization. Platform work in the open-source repo includes proxying master requests to internal services, a new AWS deployment type, deployment health checks, and WebUI integration. Go, Python, Kubernetes, AWS, TypeScript

Portable Browser - Control plane that runs remote Chromium sessions and hands agents scoped CDP access (2026)

Control plane for remote browser sessions. Workers dial out over a single persistent WebSocket, so they need no inbound ports, and that connection multiplexes control messages, Chrome DevTools Protocol tunnels, and HTTP and WebSocket proxy frames under a zod-typed wire protocol. Consumers get short-lived scoped tokens, persistent browser profiles, and shared or exclusive session locks through an OpenAPI API, generated SDKs, a CLI, and an operator UI. TypeScript, PostgreSQL, WebSocket, Chrome DevTools Protocol
live

NetsBlox - Networked block-based programming language for teaching distributed computing (2017 - 2019)

Open-source, web-based visual programming environment that extends Snap! with message passing and remote procedure calls, so students build networked apps with blocks. As a research assistant at Vanderbilt ISIS from 2017 to 2019, worked on the Node.js server: the RPC framework, service caching, the message router, API doc generation, and CI. Landed 213 merged PRs and co-authored papers on the platform at SIGCSE and IEEE Blocks and Beyond. JavaScript, Node.js, MongoDB, WebSockets, Snap!
github | live

zmenu - Fast native fuzzy menu for macOS, written in Zig directly against AppKit (2023 - 2026)

dmenu-style fuzzy menu that drives AppKit from Zig through the Objective-C runtime. Items arrive on stdin or live over a versioned, length-prefixed JSON IPC socket via zmenuctl, are filtered with fuzzy matching and a Levenshtein fallback, and the selection prints to stdout. Startup latency is measured with benchmarks that send real keyboard events, and it serves as the UI for ComboSwitcher. Started in 2023 as GMenu, a Go version, and rewritten in Zig in 2026. Zig, AppKit, Objective-C runtime, Unix sockets, Swift
github | product page

gopm - PM2-style process manager in Go with a user-level daemon and sleep-aware cron (2025 - 2026)

Process manager for long-running apps, shipped as one static Go binary. A Cobra CLI talks over a Unix socket to a user-level daemon that needs no sudo, adopts running children, persists run history in BoltDB, applies restart policies, captures logs, and sends crash-loop alerts. Cron scheduling follows wall-clock time across machine sleep, and the daemon is tuned for laptops with a single shared ticker for adopted processes. Go, Cobra, BoltDB, Unix sockets, launchd

Generalizable Q-Learning Framework - Deep Q-learning framework that trains agents for Snake, Tic-Tac-Toe, Connect-N, and CartPole (2018)

Python framework built around a reusable Double DQN agent with a target network, replay memory, Huber loss, and epsilon decay, plus a shared game interface so new environments plug in as small drivers. Supports self-play training for two-player games and ships trained Tic-Tac-Toe weights and Snake reward curves. Python, Keras, TensorFlow, OpenAI Gym, Matplotlib
github

Project Archive

Roboscape Device Provisioning (2018)

A set of solutions for batch provisioning of heterogeneous robots used for provisioning Roboscape robots. C, JS, Vue.js

Visual Machine Learning (2018)

Machine Learning and Reinforcement Learning abstractions for Snap visual programming environment. JavaScript, Tensorflow.js

Fog Alert (2018)

A 3-component fog based, intrusion detection system with basic threat detection mechanisms. Using a Raspberry Pi Zero, a beefy RPi 3, and NodeMCU modules Python, C

Roboscape (2018)

Educational robotics platform for teaching fundamental cybersecurity concepts using robots and NetsBlox C, JavaScript

NetsBlox Player (2018)

A mobile app to run networked, user-created, NetsBlox projects on iOS and Android. JS, Cordova, Ionic

Falcon9 Simulation (2018)

Numerical simulation and feasibility analysis of a reusable rocket Python, Matlab

Sketch Adventures (2018)

A game made for Global Game Jam 2018 to learn Rust. It is a game that allows users to move with the arrow keys and draw with the mouse. User drawings are classified and come to life with different behaviors to help pass the levels. Watch our demo here or see the source. Rust, Python

Teachable Robots (2017)

Accessible RPI powered robotics for the classroom (CTE paper) JS, Python

Next Block Prediction (2017)

Predicting and suggesting next blocks in block programming languages given the current flow of student's project using N-grams, Word2Vec and RNNs. Python, Tensorflow

Distributed PubSub Broker (2017)

Led a team of three to design and implement three different architectures of a resilient PubSub broker with QoS using ZeroMQ, CHORD, and Zookeeper from scratch. Python

Palantiri Gazing App (2017)

A toy app to demonstrate and exercise different concurrent programming paradigms and frameworks in Java. Java, Android, Concurrent Programming

Smart Home Automations w/ RPI & ESP (2016-P)

A collection of DIY smart home projects: hot-word detection, a re-purposed Google Assistant, Kodi voice control, smart blinds, sound-reactive LED lights, and sensor nodes. Python, C, Bash

UniDB (2016)

University Database is a crowdsourced database of universities and colleges in the states. In need of an accurate, comprehensive, and free source of information, I developed and published the platform as I was going through the graduate school application process. RoR, Ruby

Professional Certifications & Training

Ruby on Rails Development; Android Programming; Linux Professional Institute Certification (LPIC) I & II; CentOS System Administration; Cisco Certified Network Associate (CCNA); Cisco Certified Network Professional (CCNP) Routing & Switching; Leadership & Soft Skills; UI/UX Design;

Selected Publications

H.Zare, M.Maróti, A.Ledeczi "Educational Robotics with a Twist", Blocks+ Workshop, SPLASH - 2018

A.Ledeczi, H.Zare, G.Stein "NetsBlox and Wireless Robots Make Cybersecurity Fun", SIGCSE - 2019 - ACM

A.Ledeczi, M.Maróti, H.Zare, et al. "Teaching Cybersecurity with Networked Robots", SIGCSE - 2019 - ACM

B.Broll, A.Ledeczi, H.Zare, D.Do, J.Sallai, P.Volgyesi, et al. "A Visual Programming Environment for Introducing Distributed Computing to Secondary Education", Journal of Parallel and Distributed Computing - 2018 - Elsevier

B.Broll, H.Zare, A.Ledeczi "Creating Engaging Science Projects with NetsBlox", Blocks and Beyond - 2017 - IEEE

B.Broll, H.Zare, D.Do, M.Misra, A.Ledeczi "You Can Teach Computer Networking in High School", Blocks and Beyond - 2017 - IEEE

N.Hutchins, T.Darrah, H.Zare, G.Biswas "A DSML for a Robotics Environment to Support Synergistic Learning of CT and Geometry", Computational Thinking Education (CTE) - 2018

Professional Activities & Community Involvement

* Technical Mentor at Vandy Hacks hackathon events
* Conducted 100+ technical interviews for engineering positions
* Mentored 10+ new engineers in machine learning infrastructure
* Active contributor to open-source projects in the AI/ML space
* Regular participant in technical conferences and industry events

Honors, Awards & Academic Achievements

* Full Graduate Scholarship - Vanderbilt University (2017-2019)
* Top Student Award - Engineering Department, Junior & Senior Years (2014, 2015)
* Top 1% National Ranking - Iranian National Universities Entrance Exam (2011)
* Technical Leadership Recognition - Multiple performance awards at HPE/Determined AI

FILES

/usr/share/doc/ari/resume.pdf

printable resume

/srv/git/hamidzr

source code and side projects

/net/linkedin/hamidzare

professional profile

/usr/bin/contact

terminal-style contact page

AUTHOR

* Email: az@arizare.com
* Works from: San Francisco Bay Area, CA and Austin, TX
* Professional Network: Open to new opportunities and collaborations