You will get a custom YOLO object detection model, optimized for real-time edge or cloud
Rising Talent
Rising Talent
Project details
I build and deploy real-time computer vision that actually ships — on edge hardware (NVIDIA Jetson, Rockchip RK3576/RK3588) and in the cloud (AWS). Most CV models work fine in a notebook, then fall apart in production. I take them the rest of the way.
I recently took a production detection pipeline from 8 FPS to 47 FPS with a 77% reduction in model size through INT8 quantization — running real-time on a low-power edge NPU. That's the optimization most projects need but few engineers can actually deliver.
Whether you need a custom YOLO model trained on your data, a full detection + tracking pipeline, or a model optimized and deployed on your edge device, I own the complete process: dataset → training → quantization → on-device inference.
You get clean, documented code that's yours to keep, daily updates, and an engineer who stays until it runs in production — not just until the demo works.
Not sure which tier fits? Message me first and I'll scope it with you.
I recently took a production detection pipeline from 8 FPS to 47 FPS with a 77% reduction in model size through INT8 quantization — running real-time on a low-power edge NPU. That's the optimization most projects need but few engineers can actually deliver.
Whether you need a custom YOLO model trained on your data, a full detection + tracking pipeline, or a model optimized and deployed on your edge device, I own the complete process: dataset → training → quantization → on-device inference.
You get clean, documented code that's yours to keep, daily updates, and an engineer who stays until it runs in production — not just until the demo works.
Not sure which tier fits? Message me first and I'll scope it with you.
Machine Learning Tools
Keras, MLflow, NVIDIA AI Platform, OpenCV, Python, PyTorch, TensorFlow, Vertex AIWhat's included
| Service Tiers |
Starter
$100
|
Standard
$200
|
Advanced
$500
|
|---|---|---|---|
| Delivery Time | 7 days | 14 days | 28 days |
Number of Revisions | 2 | 4 | 6 |
Number of Model Variations | 1 | 2 | 3 |
Number of Scenarios | 1 | 2 | 3 |
Number of Graphs/Charts | 1 | 2 | 3 |
Model Validation/Testing | |||
Model Documentation | |||
Data Source Connectivity | - | - | |
Source Code |
Frequently asked questions
2 reviews
(2)
(0)
(0)
(0)
(0)
This project doesn't have any reviews.
AJ
Amit J.
Nov 11, 2025
Yolo based Detection for Traffic signs
RajaAli was extremely responsive and delivered great quality output for this assignment.
JC
Jean C.
Sep 18, 2025
Build 1-Page Website with Airtable & Tally Integration
Very talented
About Raja Ali
Computer Vision | Deep Learning | Edge + Cloud | YOLO, PyTorch, AWS
100%
Job Success
Lahore, Pakistan - 5:38 am local time
⚡ Recent results:
→ Model too slow for edge → 8 FPS to 47 FPS (5.9x), 77% smaller via INT8 quantization
→ Vehicle monitoring → live fleet tracking across 50+ trucks, 24/7
→ High false-positive rate → 73% fewer false alarms
→ Expensive cloud pipeline → $2,100 to $850/month
🔹 What I build:
- Object detection, tracking & counting (YOLO v5–v11, ByteTrack, DeepSORT)
- Image segmentation, pose estimation, OCR & image classification
- Edge deployment & model optimization (TensorRT, RKNN, ONNX, DeepStream, INT8/FP16)
- Cloud CV pipelines & inference APIs (AWS SageMaker/Lambda, FastAPI, Docker, Triton)
- Quality inspection, defect detection & anomaly detection
🏭 Industries:
Security & surveillance • Traffic & ANPR • Manufacturing & industrial inspection • Retail analytics • Fleet & driver monitoring • Agriculture
🛠 Stack:
PyTorch, TensorFlow, YOLO, OpenCV, TensorRT, RKNN, ONNX, DeepStream • NVIDIA Jetson (Nano/Xavier/Orin), Rockchip RK3576/RK3588 • AWS, FastAPI, Docker, Roboflow • Python, C++
✅ How I work:
I understand the problem before I build. Daily updates, no disappearing. Clean, documented code that's yours to keep. I stay until it runs in production — edge or cloud.
Need a slow model made fast, or a CV system that has to ship? Message me — let's scope it.
Steps for completing your project
After purchasing the project, send requirements so Raja Ali can start the project.
Delivery time starts when Raja Ali receives requirements from you.
Raja Ali works on your project following the steps below.
Revisions may occur after the delivery date.
Scope & Data Review
I review your data, detection classes, and deployment target, then confirm scope, accuracy goals, and timeline before any work starts.
Model Training & Tuning
Tuning train and fine-tune a custom YOLO model on your data, iterating on augmentation and hyperparameters to hit your target accuracy (mAP).