The reach behind the work
Across LinkedIn and X over the past year, the computer vision content drew 4.09 million impressions and 83,000 engagements. Use the switch below to see it all together or one platform at a time.
Reporting window: 30 Jun 2025 to 29 Jun 2026 (365 days).
Overall · LinkedIn + X
Reach over the year
Monthly impressions. Hover any month for the exact numbers.
| Jun 25 | 13,061 | 1.59% |
| Jul 25 | 483,989 | 1.91% |
| Aug 25 | 288,887 | 1.72% |
| Sep 25 | 428,623 | 2.01% |
| Oct 25 | 243,298 | 1.9% |
| Nov 25 | 274,032 | 2.19% |
| Dec 25 | 167,151 | 2.23% |
| Jan 26 | 208,765 | 1.97% |
| Feb 26 | 335,935 | 2.12% |
| Mar 26 | 146,227 | 1.97% |
| Apr 26 | 149,826 | 1.73% |
| May 26 | 194,412 | 1.51% |
| Jun 26 | 130,466 | 1.84% |
| Jun 25 | 320 | 1.25% |
| Jul 25 | 40,745 | 3.66% |
| Aug 25 | 106,950 | 2.34% |
| Sep 25 | 7,616 | 1.77% |
| Oct 25 | 7,299 | 2.97% |
| Nov 25 | 7,730 | 2.9% |
| Dec 25 | 2,543 | 3.58% |
| Jan 26 | 2,452 | 3.87% |
| Feb 26 | 1,004 | 1.1% |
| Mar 26 | 194,236 | 2.43% |
| Apr 26 | 122,886 | 2.63% |
| May 26 | 356,896 | 1.83% |
| Jun 26 | 177,631 | 2.59% |
Platform breakdown
Best days to publish
Followers work at
Audience
By seniority
By role
By industry
Top locations
LinkedIn reports top cities only.
Top posts
- 1TorchScript inference speed testSep 2025· View on LinkedIn184.9K views
- 2Ultralytics YOLOE-26 supportFeb 2026· View on LinkedIn180.8K views
- 3CoreML: YOLO11 on AppleAug 2025· View on LinkedIn131.0K views
- 4YOLO26 ONNX: +75% speedJan 2026· View on LinkedIn116.5K views
- 5MediaPipe vs Ultralytics YOLO PoseJul 2025· View on LinkedIn87.9K views
- 6Real-time pothole detectionFeb 2026· View on LinkedIn71.5K views
What the content looks like
It is hands-on computer vision: real models running on real footage, with code and numbers. This is the full range the feed covers, not a shortlist, so you can find your own category rather than guess whether it ever comes up.
Object detection
Custom classes trained on real footage, running live and holding up on busy scenes.
Object tracking
Identities held across frames, occlusions and re-entries, with counting and zone logic on top.
Instance segmentation
Pixel masks instead of boxes, for measuring area, cutting objects out and redaction.
Pose estimation
Skeleton keypoints on live video, for posture, ergonomics, rep counting and safety rules.
Image classification
Whole-image labelling by category, condition or defect, including the rare long tail.
Depth estimation
Monocular depth from an ordinary camera: distance, separation and rough 3D layout.
OCR & document AI
Text, fields and tables pulled out of invoices, IDs and forms, scans and photos included.
Video analytics
Camera feeds turned into numbers: line crossings, occupancy, dwell time and flow.
Quality inspection
Defect detection on the line: scratches, misalignment and missing parts, cost of a miss weighed.
Anomaly detection
Modelling what normal looks like, then flagging the rare events nobody has enough examples of.
Datasets & labelling
Labelling policy, hard cases first, and audits that surface broken images and split leakage.
Training & tuning
Per-class metrics and the actual failure cases, so accuracy work targets what is really wrong.
Model releases & benchmarks
First-look reactions and reproducible speed numbers for new YOLO and vision models.
Edge deployment
ONNX, TensorRT and CoreML onto Jetson, Apple and accelerators like MemryX, inside a latency budget.
Cloud & on-prem serving
The model shipped where the data is allowed to live, behind an API other systems can call.
Monitoring & retraining
Watching accuracy after launch, catching drift as conditions change, retraining on what came in.
Applied demos
Potholes, number plates, PPE, trajectory forecasting, fire and smoke. Vision on real problems.
Tutorials & teaching
Hands-on guides and code people can actually run, on LinkedIn, X and the blog.
Co-marketing partnerships
Your product in front of an audience generating over 4 million impressions a year, shown as an honest, hands-on demo rather than an ad. What moves the scope most is how much of the building is mine. Treat the list as a menu, not a fixed set.
- LinkedIn posts paced across my strongest reach window, Tuesday to Thursday
- The hook and caption rewritten in my own voice, and you review before it goes live
- Development on my side: your product wired into a real detection, tracking or segmentation workflow
- Real numbers from that work, measured on my own hardware, not taken from your deck
- Runnable code or a demo your own team can verify, so nothing rests on my word alone
- A blog on rizwanai.com or Medium, where it keeps earning traffic long after the posts stop
- A mix of formats: a live demo, a before and after, and a hard benchmark number
- A full recap of reach, engagement and the top comments, plus the raw assets for your own docs
Sponsorship FAQ
- Who actually sees a sponsored post?
- On LinkedIn, mostly working software, AI and data engineers, from entry level up to senior, with a strong base in India and Pakistan and a long tail across Europe and the US. On X the audience skews 25 to 44 and US, Japan, India and Pakistan. They work everywhere from small startups to Microsoft, NVIDIA and Amazon.
- What does a sponsored post look like?
- It looks like everything else I post. A real, hands-on look at your product inside a computer vision workflow, like a benchmark, a quick demo or a short build. It reads as content, not an ad, and that is exactly why people stop to watch.
- Do you only say positive things?
- No. The reviews and benchmarks are honest, because this audience can spot hype instantly. We agree what a post will cover before we start, but the opinion stays mine. That credibility is the whole reason a placement here is worth it.
- How much involvement can I ask for?
- Anywhere from distribution only, where the content and the ideas are yours and I put them in front of 50,000+ working computer vision engineers, through to me building with your product and benchmarking it myself across a run of posts plus a blog. It comes down to how much proof you need. Models, APIs and edge hardware usually want the hands-on end, because their claims are the kind that get checked. Tell me what you are building and I will suggest the shape.
- What does development involved actually mean?
- It means I sit down and build with your product the way any engineer would: wire it into a real detection, tracking or segmentation workflow, run it on real footage, and measure it on my own hardware. Whatever comes out of that is what gets posted. It is the difference between me telling my community your product is fast and me showing them the FPS I got. It is also the main thing that moves the scope and the cost of an engagement.
- What does it cost?
- It depends on how much of the build is mine, so I quote per engagement rather than publishing a rate card. Distribution, where the content is already yours, sits well below an engagement where I am integrating and benchmarking your product across several posts. Tell me what you are building and I will come back with a scope, a timeline and a price.
- How do we start?
- Hit Get started and tell me what you are building. I read every enquiry myself and reply within one business day with a suggested scope, a timeline and a price, plus an honest answer on whether this is a fit at all. No pressure to commit.