AIML Engineer — Computer Vision · Edge AI

Building simple systems for a complex world.

I'm Aryan Chougule — I build and ship AI systems, from 40k-image datasets to models running in real time on NVIDIA Jetson edge hardware, and lunar-surface analysis pipelines built with SAM 3, DINOv2, ViT, and MCP-driven agentic workflows.

Get in touch → Résumé ↓
scroll ↓
PyTorch✳Object Detection✳NVIDIA Jetson✳Docker · Kafka✳Vision Transformers✳RAG · Agentic AI✳ PyTorch✳Object Detection✳NVIDIA Jetson✳Docker · Kafka✳Vision Transformers✳RAG · Agentic AI✳
0
images curated & labelled
0
less manual annotation work
0
face-recognition accuracy
0
mAP@50 on detection models
( About )

I'm a B.Tech (CSE — AIML) graduate and AIML intern who likes the whole loop: building datasets, training and validating models, and deploying them to the edge where they have to run fast and stay reliable.

📍 Pune, India Open to Bangalore Graduated June 2026

Experience

01
NOV 2025 — PRESENT

AIML Intern

Ador Powertron · Pune
  • →Built and deployed real-time traffic-monitoring AI on Azure and NVIDIA Jetson edge devices (Nano, Xavier, Orin) with automated, Kafka event-streamed violation reporting — using Python, deep learning, Linux & Docker.
  • →Built an automated image-labelling tool with PyTorch and Meta AI's SAM 3, cutting manual annotation work by 70%.
  • →Delivered two object-detection models (>75% mAP@50, >60% mAP@50–95) and two instance-segmentation models (>70% mAP@50) on a self-curated 40,000+ image dataset.
  • →Managed the full annotate → train → validate → retrain cycle with Roboflow, SAM 3 & PyTorch, lifting mAP@50 with every pass.

Selected Projects

02
Live demo ↗
boulder / crater detection

Moon Forge

AI lunar-surface analysis that processes satellite imagery to detect boulders & craters and chart safer rover paths — SAM 3 segmentation, DINOv2 depth, MCP-driven, Flask + Vercel CI/CD.

Python MCP SAM 3 DINOv2 Flask
GitHub ↗
vector embedding retrieval

Facial Recognition · RAR

A face-recognition system using RAG-style vector embedding retrieval (ArcFace + MTCNN, cosine similarity) — 96%+ accuracy from just 10 enrolment images per person.

TensorFlow ArcFace MTCNN RAG Vector DB

Toolkit

03
AI / Machine Learning
PyTorch · TensorFlow · Keras · Scikit-learn · Pandas · NumPy · LLMs · RAG · Agentic AI · GenAI
Computer Vision
Object Detection · Semantic Segmentation · VLMs · Vision Transformers (ViT) · BotSORT · OpenCV · Embedding Retrieval
Deployment & MLOps
Docker · Kubernetes · FastAPI · ONNX / TensorRT · NVIDIA Jetson · FFmpeg · Twingate
Languages & Databases
Python · C++ · MySQL · PostgreSQL · MongoDB · SQLite · Vector Databases
Developer Tools
Git · Claude Code · OpenAI Codex · Roboflow · Confluent Kafka · Streamlit · Postman · Ubuntu · Hugging Face
Education
B.Tech CSE (AIML)
Rajarambapu Institute of Technology · CGPA 7.87 · 2026
Supervised Machine Learning — Stanford Online / DeepLearning.AI (Coursera)
( Let's talk )

Have a problem worth
solving?

aryanchougule4747@gmail.com +91 93734 47919
LinkedIn ↗ GitHub ↗ Résumé PDF ↓ Pune, India — --:-- IST
© 2026 Aryan Chougule — Built with intent.