Machine Learning

Train metric-learning embeddings for retrieval-based product recognition

Train NVDINOv2 vision backbones with label-free self-distillation

Train NVPanoptix3D for panoptic 3D scene reconstruction from posed RGB images

Train OCDNet models that detect arbitrary-oriented text in natural images

Train OCRNet models that read text in cropped images with CTC or attention

Train OneFormer for universal segmentation with one task-conditioned model

Train Siamese optical inspection models that flag defects from image pairs

Train PointPillars models for 3D object detection from LiDAR point clouds

Train ST-GCN pose classifiers that label action categories from skeleton sequences

Train person re-identification models that match people across camera views

Train, distill and quantize RT-DETR real-time object detectors in TAO

Train lightweight SegFormer models for real-time semantic segmentation

Run a standard train, evaluate and export workflow for any TAO model

Prepare multi-camera data and train Sparse4D models for 3D detection and tracking

Train Visual ChangeNet models for PCB defect checks and change-segmentation masks

Check TAO DAFT datasets for structure, schema and cross-reference errors

Port a paper's official CV code and recipe to your data and debug bad transfers

Patch TileGym GPU kernels into Hugging Face transformers models before loading

Select, fine-tune and deploy ML models on Amazon SageMaker

Detect anomalies in time series with the Azure AI Anomaly Detector SDK in Java

Manage Azure ML workspaces, jobs, models, datasets and pipelines in Python

Create Weights and Biases instances and manage SSO through Azure in .NET

Forecast market time series with the Kronos model and news-aware adjustments

Organize PyTorch training with LightningModules, Trainers and multi-GPU strategies

Predict molecular ADMET properties and generate molecules with graph neural networks

Search, download and upload Hugging Face models and datasets with the hf CLI

Run llama.cpp inference, convert GGUF models and quantize them for local deployment

Forecast long time series with GRU neural networks

Cluster customers or entities with K-Means to build profiles

Build decision trees to surface the key rules behind a target variable

Orchestrate GPU provisioning and training or inference jobs across GPU clouds

Create and manage portable dstack presets for model inference tuning

Prototype model serving with vLLM or SGLang on real GPUs before scaling

Compare candidate ML routes after survey, pick one and record rejected options

Run classical baselines under the same protocol and write a reproducible summary

Run the experiment stage of the Scientify research pipeline for ML models

Implement research ideas as ML model code in the Scientify research pipeline

Fine-tune and run models on Tinker with LoRA, SFT and reinforcement learning

Manage algorithm experiments and visualize results to tune complex models

Run visual recognition and inspection for industrial sites like energy and ports

Build decision tree models in R and export feature importance tables and figures

Select predictive genes with elastic net logistic regression and cross-validation plots

Generate, predict and embed protein sequences with ESM3 and ESM C models

Validate a prognostic risk signature on an external cohort with KM and ROC plots

Train embeddings on genomic interval BED files for chromatin accessibility studies

Build LASSO logistic models on omics features with cross-validation and coefficient paths

Train LightGBM models in R and export metrics, feature rankings and importance plots

Match a medical research problem to recent algorithms and method papers

Plan non-oncology diagnostic machine learning studies with feature selection