NVIDIA profiles startups using vision AI to sort mixed waste
Four Inception companies run NVIDIA edge GPUs to identify plastics, packaging and textiles on fast conveyor belts.
1
Stories per day · 30 days
1 RSS
Relatednvidia4hardware-security2nemotron2qwen2robotics2adversarial-patches1agents1anomaly-detection1bolt1co-design1cold-boot1computer-vision1cyber-physical-systems1endoscopic-segmentation1flood-detection1flox-intelligence1gguf1image-classification1
30 stories, newest first · one card per story
Stories tagged #edge-ai
Underdog demos 32MB Woof speech model for edge devices
The 32MB Woof model runs locally in browsers and supports typing to speech without internet.
1
Software PUFs surveyed for hardware roots of trust in edge AI
Survey of 37 software PUF works highlights trade-offs for boot-time key derivation and response latency in edge AI.
HydroReasoner advances floodwater detection with knowledge-guided reasoning
HydroReasoner uses Qwen2.5-VL and a floodwater knowledge graph to improve reasoning and generalize to varied scenes while running efficiently on edge devices.
Stacked hardware controller enhances Edge AI security
A stacked hardware security controller enables low-overhead anomaly detection in Edge AI systems using performance counters.
Researchers propose post-quantum key rotation for edge and physical AI agents
PARS-ESL records key epochs and membership, not operational traffic, so constrained agents can act as lightweight clients.
Signed knowledge packs miss poisoned offline RAG updates on Pi 5
A validly signed poisoned pack zeroed 30 previously correct answers, and 14 returned the injected answer; hash checks missed six of nine attacks.
Signed knowledge packs still accept poisoned updates in offline edge RAG
On an isolated Raspberry Pi 5, a validly signed poisoned pack drove correct answers on 30 questions to zero.
Unsloth publishes GGUF of NVIDIA Nemotron 3.5 Bolt E4B
The Q4_K_M build has 8B total parameters, 4B effective, and a 64K-token context.
TernScope co-designs ternary models for efficient segmentation
TernScope delivers a 0.2483M parameter model with 88.46% lower compute area than prior accelerators while retaining accuracy.
Edge vision-language models match detectors on physical patch evasion
A unified metric finds compact grounding VLMs can fail with confident wrong boxes, not only missed people.
IndustrialRemembr extends NVIDIA ReMEmbR to robot fleets
Robots encode video on Jetson devices and generate scene descriptions only on demand, without uploading raw frames.
Open From Data Center to Desktop: NVIDIA Expands Nemotron Family With Highest Efficiency Super VL and Bolt Models
NVIDIA released open Nemotron 3.5 Super VL, a 120B agent model, plus Bolt edge models.
When Rotation Bounds Nothing: Streaming Image Classifiers to Field Nodes Under Physical Capture
Rotating streamed image classifiers fails to stop model recovery after physical seizure.
EC2-Bench evaluates edge-cloud collaboration for long-horizon LLM agents
Collaboration beat edge-only and cloud-only on performance, resource use, and privacy, while privacy stayed strategy-dependent.
LiquidAI/LFM2.5-24B-A2B-GGUF — new model trending #30 on Hugging Face
Liquid AI released GGUF weights for LFM2.5-24B-A2B, including a Q4_0 distillation build.
Protecting CPU AI On Edge TEEs: WebAssembly's Promise and Practical Challenges
OP-TEE runs unaltered WebAssembly AI models on Arm TrustZone with 22% overhead and 6% extra latency.
Perplexity AI Releases pplx-embed-v2-late: A 0.6B Edge Model and a 9B Model Scoring 92.4% on MADQA
Perplexity released pplx-embed-v2-late, a 0.6B edge embedder and a 9B model scoring 92.4% on MADQA.
Cactus-Compute/whistle — new model trending #30 on Hugging Face
Cactus Compute released Whistle, a 16.9 MB on-device speech-to-text model for seven languages without a GPU.
4
EdgeCraft: Automated Model Crafting for Edge IoT
EdgeCraft uses an LLM to craft deployable edge ML models that meet quality, latency, and energy goals.
LightMIS: Ultra-Lightweight Medical Image Segmentation Without a Stage-Wise Decoder
LightMIS segments medical images with 0.131M parameters and near Mobile U-ViT Dice while cutting compute sharply.
TERMon: Detecting Persistent Behavioral Threats in Edge AI via Hardware-Native Ternary Runtime Monitor
TERMon, an FPGA runtime monitor, flags harmful weight corruption in edge AI using ternary behavioral fingerprints.
Small AI models let drones autonomously identify and attack battlefield targets
NATO-backed startup Scaleout Systems uses federated learning and small edge AI models to let military drones autonomously identify, prioritize, and attack battlefield targets.
2
QoS-Aware Federated Learning for Multimodal In-Cabin Interaction in Smart Vehicles
Researchers propose FedQoS, an asynchronous event-triggered federated learning framework for smart vehicles that cuts communication overhead 76.7% and latency 26.0%.