Arian Boukani

Who I am

About

Hey there! I'm a full-stack software engineer who likes owning things end to end — frontend, backend, and the cloud infrastructure underneath. I recently wrapped up my MSc in Computer Science at York University, where I built and open-sourced OptiServe, a Python framework that tunes serverless AI/ML workflows for cost, latency, and accuracy.

Arian Boukani

Software Engineer — Backend & Full-Stack Systems

I've shipped products with React/Next.js, FastAPI, .NET, and Postgres, deployed them on AWS and GCP with CI/CD, and built multi-agent AI systems that route work across specialized model pipelines. I pick up unfamiliar tools quickly under real deadlines — most recently by shipping production-style demos in 24–48 hour hackathons.

  • City: Toronto, Canada
  • Education: MSc Computer Science, YorkU
  • Currently: Software Engineering Data Trainer at Cohere
  • Interests:
    • Backend & Distributed Systems
    • Serverless & Cloud Platforms
    • Full-Stack Product Engineering
    • Systems for Machine Learning
  • Age:
Download Résumé Get in Touch
200+

Monthly active users on SevenTask

99.1%

Profiling accuracy against live AWS executions

89.6%

Of 1,250 problems solved within 5% of optimal

3

Papers published or under review

Where I've been

Résumé

A brief description of what I've done till now.

Experience

Software Engineering Data Trainer

Jul 2026 - Present

Cohere, Part-Time Contract — Toronto, Canada (Remote)

  • Evaluate, annotate, and code-review AI-generated responses across Python, Java, SQL, JavaScript, and HTML for correctness and strict adherence to task instructions
  • Compare competing model responses and write detailed feedback to improve coding practices and solution quality
  • Apply algorithm design knowledge to craft accurate reference solutions for complex coding prompts

Research Assistant — Software Engineering, ML Systems

Sep 2023 - Jul 2026

PACS Lab, York University, Full-Time — Toronto, Canada

  • Designed, built, and open-sourced OptiServe, a Python framework that auto-tunes AWS Lambda + Step Functions workflows across latency, cost, and accuracy; owned API design, packaging, documentation, and CI
  • Implemented profiling, performance and cost models, and benchmarking infrastructure validated to 99.1% accuracy against live AWS executions and 99.89% against AWS billing data across 3,000 warm invocations per workflow — enabling what-if analysis across memory tiers (128 MB–10 GB) and model variants with zero redeployment or re-profiling
  • Delivered optimization heuristics for an NP-hard configuration-search problem, coming within 5% of the optimal configuration in 89.6% of 1,250 evaluated problems across six serverless workflows spanning ResNet, YOLO, and BERT model variants
  • Established the project's engineering workflow: unit and integration test suites, GitHub Actions CI, containerized reproducible environments (Docker), and code review with semantic versioning to keep the open-source API stable across releases

Co-Founder, Software Engineer

Jul 2021 - Dec 2022

SevenTask, Full-Time — Remote

  • Co-founded a task-management SaaS; designed, built, and deployed the product on a containerized stack, growing it to 200+ monthly active users
  • Architected and built core features end-to-end in Angular + .NET: a Kanban board, calendar, JWT-based authentication, WebSocket-powered real-time chat, and Redis-backed caching
  • Led a team of 5 engineers through Kanban v1 delivery; ran sprint planning, code review, and pull-request workflows
  • Designed and documented REST APIs with Swagger/OpenAPI; instrumented A/B tests and feedback loops to validate releases
Watch the demo

Education

Master of Science in Computer Science

Sep 2023 - Feb 2026
York University

York University, Toronto, Canada

  • Research assistant at the PACS lab under the supervision of Prof. Khazaei
  • Research focus: serverless ML systems, cloud performance modeling, and cost-latency-accuracy optimization
  • CGPA: 4 / 4

Bachelor of Science in Computer Engineering

Sep 2018 - Aug 2023
Amirkabir University of Technology

Amirkabir University of Technology, Tehran, Iran

  • Recognized as an outstanding student
  • CGPA: 18.60 / 20 (3.98 / 4)

Diploma in Mathematics and Physics

Sep 2014 - Jul 2018
NODET

Shahid Beheshti High School (NODET), Mahabad, Iran

  • CGPA: 4 / 4

Teaching

Teaching Assistant

Feb 2022 - Dec 2025
  • York University (2023–2025) — Distributed Systems, Databases, Operating Systems, Object-Oriented Programming
  • Amirkabir University of Technology (2022–2023) — Cloud Computing, Real-Time Embedded Systems, Internet of Things
What I've built

Selected Projects

An open-source research framework and two hackathon products, each shipped end to end.

OptiServe

Open source · PACS Lab

A Python framework that auto-tunes serverless AI/ML workflows across cost, latency, and accuracy. Profile a workflow once, then answer what-if questions across memory tiers and model variants with zero redeployment or re-profiling.

99.1%Accuracy vs. live AWS executions
99.89%Cost accuracy vs. AWS billing data
89.6%Of 1,250 problems within 5% of optimal
  • Sweeps AWS Lambda + Step Functions configurations from 128 MB to 10 GB, benchmarked across six workflows spanning ResNet, YOLO, and BERT variants
  • Published as a documented, semantically versioned package with unit and integration test suites, GitHub Actions CI, and Docker-based reproducible environments
  • Python
  • AWS Lambda
  • Step Functions
  • PyTorch
  • Docker
  • GitHub Actions
View code

BakeryPilot

Top-4 finalist · TMLS hackathon

A full-stack supply-chain copilot for bakeries, built at the Toronto Machine Learning Summit.

  • A multi-agent LangGraph service routing to 5 domain agents, with human-in-the-loop action cards gating every commit
  • REST and SSE contracts between a Next.js 15 frontend and a FastAPI backend on PostgreSQL with pgvector
  • Containerized with Docker on a GCP VM, behind a CI/CD pipeline that ships every push to a live demo
  • Next.js 15
  • FastAPI
  • LangGraph
  • PostgreSQL
  • pgvector
  • SSE
  • GCP
View code

Guardian

NVIDIA Spark Hack · Toronto

An on-device home-emergency assistant for elderly care that keeps every byte of health data at the edge.

  • A multi-agent FastAPI + LangGraph backend routing to 6 specialist services, with a weighted risk monitor scoring live vitals into severity tiers
  • A Twilio integration that automatically places a 911 call on a confirmed fall
  • A real-time voice pipeline (faster-whisper, LangGraph, Kokoro TTS) streamed over SSE to a Wear OS watch, running on TensorRT on an NVIDIA DGX Spark
  • FastAPI
  • LangGraph
  • TensorRT
  • Next.js
  • Wear OS
  • Twilio
View code
My toolkit

Skills

The languages, tools, and practices I work with day to day.

Languages

  • Python
  • JavaScript
  • TypeScript
  • Go
  • Java
  • C / C++
  • SQL

Backend & APIs

  • REST
  • Microservices
  • Event-Driven Architecture
  • FastAPI
  • .NET
  • Swagger / OpenAPI
  • WebSockets
  • SSE
  • Auth (JWT)

Cloud & DevOps

  • GCP
  • AWS Lambda
  • Step Functions
  • S3
  • CloudWatch
  • Docker
  • Kubernetes
  • GitHub Actions
  • Linux
  • Git

Databases & Caching

  • PostgreSQL
  • pgvector
  • Redis
  • Schema Design
  • Indexing
  • Query Optimization

CS Foundations

  • Data Structures
  • Algorithms
  • Concurrency
  • Distributed Systems
  • Operating Systems
  • Object-Oriented Design
  • Design Patterns
  • System Design

Testing & Practices

  • Unit Testing
  • Integration Testing
  • End-to-End Testing
  • Code Review
  • Debugging
  • Profiling
  • Observability
  • Agile / Scrum
  • A/B Testing

Frontend

  • React
  • Next.js
  • Angular
  • HTML
  • CSS

AI-Assisted Dev

  • Claude Code
  • ChatGPT
  • Prompt Engineering
  • Reviewing AI-Generated Code
  • Agentic Workflows

Applied ML

  • PyTorch
  • LangGraph
  • Pandas
  • NumPy
Research

Publications & Talks

Peer-reviewed work on performance prediction and serverless ML systems.

ACM/SPEC ICPE 2025

PreNeT

Leveraging Computational Features to Predict Deep Neural Network Training Time.

Read paper View code
TMLR

CAPE

Generalized Convergence Prediction Across Architectures Without Full Training.

Read paper View code
Under review · IEEE TPDS

OptiServe

Cost-Aware, Performance-Driven, and Accuracy-Tuned Serverless Applications.

View code

Talks

Industry Night — York University, 2025

Presented a poster on PreNeT to engineers from AMD, Telus, and Dell.

Awards

Top-4 Finalist

TMLS Toronto ML Summit Agentic Hackathon, for BakeryPilot.

Outstanding Student Award

Amirkabir University of Technology, during the BSc in Computer Engineering.

Say hello

Contact

Always happy to talk about backend and distributed systems, serverless platforms, or a project you're working on.

Let's build something together.

The fastest way to reach me is email — I read everything and reply to most of it.

Email me Download Résumé