ARYA GOSAVI
ABOUTPROJECTSTIMELINECONTACT
RESUME
TORONTO, CASOFTWARE / AI / HARDWARE

ARYAGOSAVI

↘
Software engineering, AI, embedded systems, and full-stack development.
SOFTWARE ENGINEER

Software engineer working across AI, embedded systems, and full-stack development. Studying at York University and building software that connects AI with real-world systems.

VIEW MY RESUMEVIEW PROJECTS ↓

ABOUT

I'm Arya, a software engineer drawn to the boundary where code meets the physical world.

I study Software Engineering in the Big Data stream at York University. My work spans AI applications, computer vision, full-stack products, and embedded systems.

At FGF Brands, I built document-understanding and retrieval tools.

HANDS-ON ENGINEERING

Student teams

Software work across rocketry and motorsports.

01CURRENT

Arbalest Rocketry

CURRENT TEAMS
  • Payload Software
  • Imaging
PREVIOUS TEAMS
  • Flight Computer Software
02TEAM CONTRIBUTION

Lassonde Motorsports

TEAM
  • Embedded Systems
INTERESTSASCII / 3D
LOADING 3D MODEL…
01 / 03GAMINGDRAG TO ROTATE ↗

PROJECTS

01 / 05

QUIZZLE

ROCKET TRACK

CAPTIONER

VISUOMOTOR LAB

INTENT LAB

QUIZZLE

QUIZZLE

QUIZZLE

ROCKET TRACK

AI / FULL-STACK

Quizzle question-generation flow, drawn from the API route and app structure
Rocket Track detector output with a tracked rocket bounding box
Question generation · source diagram
  • OpenAI generates structured quizzes with multiple-choice and open-ended question formats.
  • Next.js manages answering questions and reviewing results within a quiz session.
  • NextAuth associates saved games and performance history with each signed-in learner.
  • Prisma and PostgreSQL persist quiz records and progress across sessions.
  • Domain-tuned YOLOv8 produces rocket bounding boxes from launch video frames.
  • SORT links detections; coasting preserves tracks through brief detection gaps.
  • Size filtering rejects smoke trails that can otherwise produce false detections.
  • GPU latency benchmarks and TensorRT exports assess inference performance.
  • S3 stores uploads; Next.js starts and polls asynchronous AWS Transcribe jobs.
  • Automatic language detection produces transcript JSON saved back to S3.
  • The editor supports transcript corrections and caption styling before rendering.
  • Browser-based FFmpeg burns subtitles into a standalone, shareable video file.
  • Python sends MediaPipe pose JSON to Unity over local UDP.
  • Body calibration precedes reaching trials with a +30° cursor-only rotation.
  • Tracking loss pauses trials; prolonged loss aborts and requeues them.
  • Session JSON and frame/trial CSVs log physical movement separately from visual feedback.
  • PyTorch classifiers learn 17 customer-service intents from 172 example phrases.
  • Four text representations share the same three-fold cross-validation protocol.
  • Accuracy: bag-of-words 60.5%, TF-IDF 57.1%, bigrams 48.0%, embeddings 47.3%.
  • Perfect training accuracy contrasts with weaker validation results on this small dataset.
  • OpenAI generates structured quizzes with multiple-choice and open-ended question formats.
  • Next.js manages answering questions and reviewing results within a quiz session.
  • NextAuth associates saved games and performance history with each signed-in learner.
  • Prisma and PostgreSQL persist quiz records and progress across sessions.
  • OpenAI generates structured quizzes with multiple-choice and open-ended question formats.
  • Next.js manages answering questions and reviewing results within a quiz session.
  • NextAuth associates saved games and performance history with each signed-in learner.
  • Prisma and PostgreSQL persist quiz records and progress across sessions.
  • OpenAI generates structured quizzes with multiple-choice and open-ended question formats.
  • Next.js manages answering questions and reviewing results within a quiz session.
  • NextAuth associates saved games and performance history with each signed-in learner.
  • Prisma and PostgreSQL persist quiz records and progress across sessions.
  • Domain-tuned YOLOv8 produces rocket bounding boxes from launch video frames.
  • SORT links detections; coasting preserves tracks through brief detection gaps.
  • Size filtering rejects smoke trails that can otherwise produce false detections.
  • GPU latency benchmarks and TensorRT exports assess inference performance.
SOURCE LIVE SITE

Next.js / TypeScript / OpenAI / NextAuth / Prisma / PostgreSQL

Python / YOLOv8 / SORT / OpenCV / TensorRT

Next.js / React / AWS S3 / AWS Transcribe / FFmpeg

Unity / C# / Python / MediaPipe / OpenCV / UDP

Python / PyTorch / NLTK / NumPy / PyTest

↓ SCROLL DOWN FOR NEXT PROJECT ↓
QUIZZLE. AI-generated quizzes. Quizzle question-generation flow, drawn from the API route and app structure
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                    

EXPERIENCE

SEP 2025AUG 2026

SOFTWARE DEVELOPER INTERN

FGF BRANDS

TORONTO, ON
  • Built an OCR and manual-review application with Flask, Kendo React, and Azure AI Foundry.
  • Connected PDF storage and structured results through Azure Blob Storage and Cosmos DB.
  • Developed MCP integrations, retrieval systems, and Azure DevOps delivery pipelines.
MAY 2025AUG 2025

AI & WEB DEVELOPMENT CO-OP

PHARMSHIFT

TORONTO, ON
  • Received a $4,000 BEST Co-op Program award for an AI-powered pharmacy OSCE preparation platform.
  • Built contextual Q&A with LangChain, retrieval-augmented generation, and the DeepSeek API.
  • Developed Next.js and Node.js features and deployed containerized services on Google Cloud.
MAR 2025AUG 2025

AI SOFTWARE DEVELOPER INTERN

GROOVEAI

REMOTE · EDMONTON, AB · PART-TIME
  • Built retail assistants with the OpenAI Agents SDK and Responses API.
  • Developed a FastAPI backend with caching for product and store-specific questions.
  • Connected a Next.js conversation interface to client product knowledge bases.
JUN 2023AUG 2023

SOFTWARE DEVELOPER INTERN

LELA SACKS AGRO

REMOTE · KANO, NIGERIA
  • Maintained a Java Spring Boot production-management system supporting 130 weaving machines and 80+ staff.
  • Improved PostgreSQL retrieval and automated testing with JUnit, Jenkins, and shell scripts.

LET'S BUILDSOMETHING GREATTOGETHER

ARYAGSV@GMAIL.COM
© 2026 ARYAGITHUB ↗LINKEDIN ↗BACK TO TOP ↑
01/05↢↣
Dark mode.
Portfolio loading.
ARYA GOSAVI / SYSTEM BOOT
       .  *  .
    .  /|\  .
   *  / | \  *
  ---<  +  >---
   *  \ | /  *
    .  \|/  .
       '  *  '

loading portfolio_

[░░░░░░░░░░░░░░░░░░░░] 0%

INITIALIZING