Computer Vision Engineer
AlterSense Limited
- Tejgaon, Dhaka
- Permanent
- Full-time
- Perform detailed analysis of computer vision models (CNNs, Vision Transformers, Video Models, etc.) at the layer/block level to understand internal representations and bottlenecks.
- Optimize model performance in terms of latency, FLOPs, memory footprint, and accuracy trade-offs.
- Conduct data adaptation and pre-processing strategies by examining the relationship between input distributions and model outputs.
- Implement model compression techniques such as pruning, quantization, and knowledge distillation for efficient deployment.
- Profile and debug models to identify performance bottlenecks across training and inference pipelines.
- Collaborate with data scientists and ML engineers to design scalable end-to-end vision pipelines.
- Research and integrate state-of-the-art architectures for tasks such as classification, detection, segmentation, or action recognition.
- Benchmark and document layer-wise behaviors for explainability, interpretability, and optimization insights.
- Master of Science (MSc) in Computer Science & Engineering
- Bachelor of Science (BSc) in Electrical & Electronic Engineering
- Master of Science (MSc) in Computer Science & Engineering
- Prior work on large-scale video understanding or vision transformer architectures.·
- Experience with low-level performance optimization (CUDA, ONNX Runtime, TensorRT).·
- Contribution to open-source projects in vision or model optimization.·
- Understanding of dataset bias, domain adaptation, and generalization techniques.
- 2 to 4 years
- The applicants should have experience in the following business area(s): IT Enabled Service, Artificial Intelligence (AI) Startup
- Age 27 to 35 years
- Only Male
- Strong background in Deep Learning and Computer Vision (e.g., CNNs, Transformers, video models).·
- Proficiency with PyTorch or TensorFlow for model training and analysis.·
- Hands-on experience in model profiling tools (e.g., PyTorch Profiler, TensorBoard, NVIDIA Nsight).·
- Solid understanding of network architectures (residual blocks, attention, 3D convolutions, etc.).·
- Experience with optimization techniques: pruning, quantization, mixed-precision training, distillation.·
- Hands-on experience with GPU environments (CUDA, cuDNN, multi-GPU/distributed training, profiling tools).·
- Strong skills in data analysis and preprocessing to align input-output distributions.·
- Familiarity with deployment constraints (edge devices, GPU inference, real-time systems).·
- Excellent problem-solving skills with an ability to think at both system and model levels.
- Performance bonus,Weekly 2 holidays
- Lunch Facilities: Partially Subsidize
- Salary Review: Yearly
- Festival Bonus: 2
Bdjobs