1.5 Key AI Use Cases and Industries
What the exam tests
Which AI capabilities apply to which industries, and which NVIDIA products/platforms enable them.
Use cases by AI capability
Generative AI
- Text generation / LLMs: Copilots, chatbots, summarization, code generation, document drafting
- Image synthesis: Product design, marketing, drug molecule generation
- Video generation: Entertainment, synthetic training data for autonomous vehicles
- NVIDIA solution: Blackwell B200/B300 (training), H100/L40S (inference); NeMo framework
Computer Vision
- Object detection / classification: Quality control in manufacturing, medical imaging, autonomous vehicles
- Video analytics: Surveillance, retail foot-traffic, sports analytics
- NVIDIA solution: NVIDIA Metropolis (smart cities/retail), Holoscan (medical), L40S/L4 for inference
Natural Language Processing (NLP)
- Semantic search: Enterprise knowledge retrieval, legal document review
- Translation: Real-time multilingual customer support
- Sentiment analysis: Financial trading signals, brand monitoring
- NVIDIA solution: NeMo, TensorRT-LLM, H100/B200
Recommendation Systems
- Collaborative filtering, deep learning ranking: E-commerce, streaming, social media
- NVIDIA solution: NVIDIA Merlin (end-to-end recommender system framework), A100/H100 for training
Autonomous Systems
- Self-driving vehicles: Perception, path planning, sensor fusion
- Robotics: Pick-and-place, warehouse automation, surgical robots
- NVIDIA solution: NVIDIA DRIVE platform (automotive), Isaac (robotics), Orin SoC
High-Performance Computing (HPC) + AI
- Scientific simulation: Climate modeling, molecular dynamics, computational fluid dynamics
- Drug discovery: Protein structure prediction (AlphaFold-class models), virtual screening
- NVIDIA solution: Grace CPU, Hopper H100, DGX SuperPOD
Industries and their AI applications
| Industry | Key AI applications | NVIDIA platforms |
|---|---|---|
| Healthcare & Life Sciences | Medical imaging diagnosis, drug discovery, genomics, robotic surgery | Clara (medical imaging), BioNeMo (drug discovery) |
| Financial Services | Fraud detection, algorithmic trading, risk modeling, AML | RAPIDS (data analytics), Morpheus (cybersecurity) |
| Manufacturing | Predictive maintenance, visual inspection, digital twins, robotics | Metropolis, Isaac, Omniverse |
| Retail & E-commerce | Recommendation engines, demand forecasting, cashier-less stores | Merlin (recommenders), Metropolis (vision) |
| Telecommunications | Network optimization, predictive maintenance, 5G RAN processing | AI-RAN, BlueField DPU for telco |
| Energy | Seismic analysis, equipment fault prediction, smart grid optimization | HPC clusters, DGX systems |
| Media & Entertainment | AI content creation, upscaling, rendering acceleration, VFX | L40S, Omniverse, DLSS |
| Government & Defense | Intelligence analysis, logistics optimization, autonomous systems | NVIDIA-Certified Systems, edge AI |
AI Factory vs AI Cloud
From the NVIDIA course, two distinct deployment models for AI infrastructure:
AI Factory:
- Purpose-built for single/few massive AI training workloads
- Runs the largest AI models (trillion-parameter class)
- Uses NVLink-based scale-up fabric + InfiniBand for scale-out
- Example: NVIDIA DGX SuperPOD, hyperscaler internal training clusters
AI Cloud:
- Hyperscale multi-tenant service hosting many diverse AI workloads
- Serves many users simultaneously with different model sizes and task types
- Uses Ethernet-based networking (more flexible/cost-efficient at cloud scale)
- Example: AWS, Azure, GCP GPU instances
Self-check questions
- Which NVIDIA platform is designed for end-to-end recommendation system development?
- What is the difference between an AI Factory and an AI Cloud deployment?
- Which NVIDIA SDK targets medical imaging AI?
- Name two AI use cases specific to the financial services industry.
- Which NVIDIA platform handles 5G RAN processing and telco AI workloads?
Answers
1. NVIDIA Merlin2. AI Factory: purpose-built for a single or few extremely large training workloads; uses NVLink + InfiniBand; designed for maximum throughput on one job. AI Cloud: multi-tenant, diverse workloads, many concurrent users/models; uses Ethernet networking; designed for flexibility and utilization across many jobs.
3. NVIDIA Clara
4. Any two of: fraud detection, algorithmic trading, risk modeling, anti-money laundering (AML), credit scoring, customer churn prediction.
5. NVIDIA AI-RAN; BlueField DPU is used to offload RAN processing in 5G deployments.