You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Seede AI helps you create a poster in 1 min. Transform ideas into professional designs in 1 click! Seede.ai's AI design generator creates social posts & marketing materials instantly
Design a family of SOTA CNN architectures (ConvNeXt/ResNeXt-style) for ESC-50 audio spectrogram classification, parameterized across 5M, 10M, 20M, and 30M sizes. The design uses mel-spectrogram inputs, depthwise separable convolutions, squeeze-and-excitation blocks, and proper regularization (SpecAugment, Mixup, Label Smoothing) to achieve strong g
Design and build a privacy-preserving hate speech detection (HSD) system that operates agnostically to author identity. The model uses a Transformer backbone (e.g., BERT/RoBERTa) with differentially private training (DP-SGD/DP-Adam) and adversarial identity disentanglement to decouple hate-signal from author-specific stylistic fingerprints, maximiz
Design a hybrid ML architecture that uses a GBDT ensemble as a hard-routed geometry extractor (producing one-hot leaf embeddings for Q/K), discards default leaf scores, and replaces them with a KRR or MLP-learned value representation (V). The final prediction is a dot-product Q·K·V aggregation mimicking a zero-temperature attention head, front-load
Design an ML-enhanced receiver chain for IoT-to-LEO-satellite communication, where conventional signal-processing algorithms (e.g., channel estimation, synchronization, detection) are augmented or replaced by learned models under low-SNR, high-Doppler, and resource-constrained IoT conditions.
StackMyAI.com is a blog that breaks down complex AI tools and SaaS platforms into simple, actionable insights. This repo includes scripts and data used to power our guides and comparisons — ensuring accuracy and transparency.
Design 'EfficientDenseNet' — a DenseNet variant under 10M params that incorporates VoVNet-style one-shot aggregation (OSA), depthwise separable bottlenecks, squeeze-and-excitation attention, LayerNorm pre-activation, and aggressive compound scaling. The goal is to train 2-3× faster than DenseNet-121 while maintaining or improving accuracy on ImageN
Design a neuro-symbolic memory architecture for autonomous agricultural agents, inspired by the brain's Complementary Learning Systems (CLS). The architecture decouples memory into a fast-learning episodic Knowledge Graph for temporal/spatial sequences (crop cycles, disease spread) and a slow-learning semantic ML layer that compresses mature data i
Design a multi-class classification and interpretability pipeline that distinguishes rheumatoid arthritis (RA), psoriatic arthritis (PsA), and osteoarthritis (OA) from hand/feet X-ray images. The architecture combines a CNN backbone (e.g., DenseNet or EfficientNet) for disease classification with an attention-based or Grad-CAM explanation module to
Design a reinforcement learning architecture for autonomous vulnerability discovery in security capture-the-flag (CTF) challenges, building on the Agent Web Model's hierarchical abstraction layers. The system will model CTF environments as Markov decision processes and develop RL algorithms (e.g., deep Q-learning or PPO with hierarchical extensions
Design ModernBERT-Pro, an enhanced encoder-only transformer incorporating Multi-head Latent Attention (MLA) for global layers, fine-grained Mixture-of-Experts FFN layers, Mamba-2 hybrid blocks for linear-complexity local mixing, and a multi-stage curriculum pretraining pipeline. The target is a ~350M-param model (with ~2.5x effective FLOPs via spar
Design a multi-shopper, multi-product autonomous checkout vision system that generalises a gesture-based prototype. The system integrates hand-item association (shopper-ID to item-ID), temporal cart-state modelling via a state machine or transformer, and robustness to occlusions, put-backs, and hand-offs. Evaluation targets on-device latency and pr
Design a diffusion-based LLM architecture that replaces autoregressive next-token prediction with iterative denoising over discrete token sequences. The architecture uses a bidirectional transformer as the denoising backbone, a continuous or discrete diffusion process over token embeddings, classifier-free guidance for conditioning, and accelerated
Design a generative model (diffusion/flow matching) for molecular geometry generation, leveraging graph neural networks (e.g., GCN, MPNN) as the backbone for representing molecular structures, with a focus on guided generation for pharmaceutical drug discovery.
Design a vision-based reinforcement learning model that enables a robotic system to detect, track, and physically intercept a fly in 3D space. The model combines a lightweight object-detection backbone for real-time fly localization with a learned control policy (e.g., PPO or SAC) that outputs motor commands to maneuver a catching mechanism (e.g.,