Deploying Tencent Hunyuan in Production: Low-Latency Inference & Cost Optimization Training Course
Deploying Tencent Hunyuan in Production: Low-Latency Inference and Cost Optimisation is a hands-on course focused on serving Tencent Hunyuan models reliably at scale.
This instructor-led live training (available online or onsite) is designed for intermediate-level engineers and architects who want to leverage Tencent Hunyuan to deploy large and Mixture of Experts (MoE) models with reduced latency, improved GPU utilisation, and controlled operational costs.
Upon completing this training, participants will be able to:
- explain the primary production challenges associated with serving Tencent Hunyuan models.
- apply practical inference optimisation techniques, including TensorRT, KV-cache tuning, quantisation, and batching.
- design a scalable deployment strategy incorporating autoscaling, monitoring, and capacity planning.
- optimise the trade-off between latency and cost for real-world production workloads.
Course Format
- Interactive lectures and discussions.
- Extensive exercises and practice sessions.
- Hands-on implementation in a live lab environment.
Course Customisation Options
- To request a tailored training session for this course, please contact us to make arrangements.
Course Outline
Tencent Hunyuan Production Fundamentals
- Overview of Tencent Hunyuan model serving scenarios.
- Production characteristics of large and MoE models.
- Common latency, throughput, and cost bottlenecks.
- Defining service-level objectives for inference workloads.
Deployment Architecture and Serving Flow
- Core components of a production inference stack.
- Choosing between containerised, on-premise, and cloud deployment models.
- Model loading, request routing, and GPU allocation basics.
- Designing for reliability and operational simplicity.
Latency Optimisation in Practice
- Using optimised inference engines such as TensorRT where applicable.
- KV-cache concepts and practical cache tuning.
- Reducing startup, warmup, and response overhead.
- Measuring time to first token and token generation speed.
Throughput, Batching, and GPU Efficiency
- Continuous batching and request batching strategies.
- Managing concurrency and queue behaviour.
- Improving GPU utilisation without compromising user experience.
- Handling long-context and mixed-workload requests.
Quantisation and Cost Control
- Why quantisation matters for production serving.
- Practical trade-offs of FP16, INT8, and other common precision options.
- Balancing model quality, latency, and infrastructure cost.
- Building a simple cost optimisation checklist.
Operations, Monitoring, and Readiness Review
- Autoscaling triggers for inference services.
- Monitoring latency, throughput, cache usage, and GPU health.
- Logging, alerting, and incident response basics.
- Reviewing a reference deployment and creating an improvement plan.
Requirements
- Basic understanding of large language model deployment and inference workflows.
- Experience with containers, cloud or on-premise infrastructure, and API-based services.
- Working knowledge of Python or system engineering tasks.
Audience
- ML engineers deploying LLMs into production.
- Platform engineers responsible for GPU-based inference services.
- Solution architects designing scalable AI serving platforms.
Custom Corporate Training
Training solutions designed exclusively for businesses.
- Customized Content: We adapt the syllabus and practical exercises to the real goals and needs of your project.
- Flexible Schedule: Dates and times adapted to your team's agenda.
- Format: Online (live), In-company (at your offices), or Hybrid.
Price per private group, online live training, starting from 2600 € + VAT*
Contact us for an exact quote and to hear our latest promotions
(*The final price may vary depending on the technical specialization of the course, the level of customization, the method of delivery and the number of learners)
Need help picking the right course?
info@nobleprog.pt or +351 30 050 9666
Deploying Tencent Hunyuan in Production: Low-Latency Inference & Cost Optimization Training Course - Enquiry
Deploying Tencent Hunyuan in Production: Low-Latency Inference & Cost Optimization - Consultancy Enquiry
Provisional Upcoming Courses (Contact Us For More Information)
Related Courses
Advanced LangGraph: Optimization, Debugging, and Monitoring Complex Graphs
35 HoursLangGraph serves as a framework for developing stateful, multi-actor LLM applications through composable graphs that maintain persistent state and provide precise control over execution workflows.
This instructor-led live training (available online or onsite) is designed for advanced AI platform engineers, AI-focused DevOps professionals, and ML architects who aim to optimize, debug, monitor, and operate production-grade LangGraph systems.
By the conclusion of this training, participants will be able to:
- Design and optimize complex LangGraph topologies to enhance speed, reduce costs, and improve scalability.
- Engineer reliability through the implementation of retries, timeouts, idempotency, and checkpoint-based recovery mechanisms.
- Debug and trace graph executions, inspect state variables, and systematically reproduce production issues.
- Instrument graphs with logs, metrics, and traces, deploy them to production environments, and monitor SLAs and costs effectively.
Format of the Course
- Interactive lectures and discussions.
- Extensive exercises and practical practice.
- Hands-on implementation within a live laboratory environment.
Course Customization Options
- To request customized training for this course, please contact us to arrange it.
Building Coding Agents with Devstral: From Agent Design to Tooling
14 HoursDevstral is an open-source framework engineered to build and operate coding agents capable of interacting with codebases, developer tools, and APIs to boost engineering productivity.
This instructor-led live training (available online or onsite) targets intermediate to advanced-level ML engineers, developer-tooling teams, and SREs who aim to design, implement, and optimize coding agents using Devstral.
Upon completion of this training, participants will be able to:
- Configure and set up Devstral for coding agent development.
- Design agentic workflows tailored for codebase exploration and modification.
- Integrate coding agents seamlessly with developer tools and APIs.
- Apply best practices for secure and efficient agent deployment.
Course Format
- Interactive lectures and discussions.
- Extensive exercises and practical application.
- Hands-on implementation within a live-lab environment.
Customization Options
- For personalized training requests, please contact us to arrange.
Open-Source Model Ops: Self-Hosting, Fine-Tuning and Governance with Devstral & Mistral Models
14 HoursDevstral and Mistral are open-source AI technologies engineered for flexible deployment, fine-tuning capabilities, and scalable integration.
This instructor-led live training (available online or onsite) targets intermediate to advanced ML engineers, platform teams, and research engineers looking to self-host, fine-tune, and govern Mistral and Devstral models within production environments.
Upon completion of this training, participants will be able to:
- Establish and configure self-hosted environments for Mistral and Devstral models.
- Apply fine-tuning techniques to enhance domain-specific performance.
- Implement versioning, monitoring, and lifecycle governance mechanisms.
- Ensure security, compliance, and responsible usage of open-source models.
Course Format
- Interactive lectures and discussions.
- Practical exercises focused on self-hosting and fine-tuning.
- Live-lab implementation of governance and monitoring pipelines.
Customization Options
- To request customized training for this course, please contact us to arrange.
LangGraph Applications in Finance
35 HoursLangGraph serves as a framework for constructing stateful, multi-agent LLM applications by utilizing composable graphs that maintain persistent state and offer precise control over execution flows.
This instructor-led live training, available either online or onsite, is designed for intermediate to advanced professionals aiming to design, implement, and operate LangGraph-based financial solutions that adhere to robust governance, observability, and compliance standards.
Upon completion of this training, participants will be equipped to:
- Design finance-specific LangGraph workflows that align with regulatory and audit requirements.
- Integrate financial data standards and ontologies into graph states and associated tooling.
- Implement reliability, safety mechanisms, and human-in-the-loop controls for critical processes.
- Deploy, monitor, and optimize LangGraph systems to ensure high performance, cost efficiency, and adherence to SLAs.
Course Format
- Interactive lectures and discussions.
- Extensive exercises and practical practice.
- Hands-on implementation within a live laboratory environment.
Customization Options
- To request customized training for this course, please contact us to arrange your needs.
LangGraph Foundations: Graph-Based LLM Prompting and Chaining
14 HoursLangGraph is a framework designed for constructing graph-structured LLM applications that facilitate planning, branching, tool usage, memory management, and controllable execution.
This instructor-led live training, available both online and onsite, targets beginner-level developers, prompt engineers, and data practitioners eager to design and build reliable, multi-step LLM workflows using LangGraph.
Upon completion of this training, participants will be able to:
- Explain core LangGraph concepts (nodes, edges, state) and when to use them.
- Build prompt chains that branch, call tools, and maintain memory.
- Integrate retrieval and external APIs into graph workflows.
- Test, debug, and evaluate LangGraph apps for reliability and safety.
Format of the Course
- Interactive lecture and facilitated discussion.
- Guided labs and code walkthroughs in a sandbox environment.
- Scenario-based exercises on design, testing, and evaluation.
Course Customization Options
- To request a customized training for this course, please contact us to arrange.
LangGraph in Healthcare: Workflow Orchestration for Regulated Environments
35 HoursLangGraph facilitates stateful, multi-actor workflows driven by Large Language Models (LLMs), offering precise control over execution paths and state persistence. In the healthcare sector, these capabilities are vital for ensuring compliance, interoperability, and the development of decision-support systems that align with clinical workflows.
This instructor-led, live training (available online or onsite) targets intermediate to advanced professionals looking to design, implement, and manage LangGraph-based healthcare solutions while navigating regulatory, ethical, and operational challenges.
Upon completing this training, participants will be able to:
- Design healthcare-specific LangGraph workflows that prioritise compliance and auditability.
- Integrate LangGraph applications with medical ontologies and standards (such as FHIR, SNOMED CT, and ICD).
- Apply best practices for reliability, traceability, and explainability in sensitive environments.
- Deploy, monitor, and validate LangGraph applications within healthcare production settings.
Format of the Course
- Interactive lectures and discussions.
- Hands-on exercises using real-world case studies.
- Implementation practice in a live-lab environment.
Course Customization Options
- To request tailored training for this course, please contact us to arrange.
LangGraph for Legal Applications
35 HoursLangGraph is a framework designed for building stateful, multi-actor LLM applications as composable graphs with persistent state and precise control over execution.
This instructor-led, live training (online or onsite) is aimed at intermediate-level to advanced-level professionals who wish to design, implement, and operate LangGraph-based legal solutions with the necessary compliance, traceability, and governance controls.
By the end of this training, participants will be able to:
- Design legal-specific LangGraph workflows that preserve auditability and compliance.
- Integrate legal ontologies and document standards into graph state and processing.
- Implement guardrails, human-in-the-loop approvals, and traceable decision paths.
- Deploy, monitor, and maintain LangGraph services in production with observability and cost controls.
Format of the Course
- Interactive lecture and discussion.
- Lots of exercises and practice.
- Hands-on implementation in a live-lab environment.
Course Customization Options
- To request a customized training for this course, please contact us to arrange.
Building Dynamic Workflows with LangGraph and LLM Agents
14 HoursLangGraph serves as a framework for composing graph-structured workflows involving Large Language Models (LLMs), enabling features such as branching, tool utilization, memory management, and controlled execution.
This instructor-led live training, available either online or onsite, is designed for intermediate-level engineers and product teams seeking to integrate LangGraph’s graph logic with LLM agent loops to develop dynamic, context-aware applications, including customer support agents, decision trees, and information retrieval systems.
Upon completing this training, participants will be equipped to:
- Design graph-based workflows that effectively coordinate LLM agents, tools, and memory.
- Implement conditional routing, retry mechanisms, and fallback protocols to ensure robust execution.
- Integrate retrieval processes, APIs, and structured outputs into agent loops.
- Evaluate, monitor, and fortify agent behavior to enhance reliability and safety.
Course Format
- Interactive lectures accompanied by facilitated discussions.
- Guided laboratory sessions and code walkthroughs within a sandbox environment.
- Scenario-based design exercises and peer review sessions.
Customization Options
- To request a customized version of this course, please contact us to make arrangements.
LangGraph for Marketing Automation
14 HoursLangGraph serves as a graph-based orchestration framework that empowers conditional, multi-step workflows involving large language models (LLMs) and tools, making it highly suitable for automating and personalizing content pipelines.
This instructor-led live training, available either online or onsite, targets intermediate-level marketers, content strategists, and automation developers looking to build dynamic, branching email campaigns and content generation pipelines using LangGraph.
Upon completing this training, participants will be equipped to:
- Design graph-structured workflows for content and email, incorporating conditional logic.
- Integrate LLMs, APIs, and data sources to enable automated personalization.
- Manage state, memory, and context across complex, multi-step campaigns.
- Evaluate, monitor, and optimize workflow performance and delivery outcomes.
Course Format
- Interactive lectures paired with group discussions.
- Practical labs focused on implementing email workflows and content pipelines.
- Scenario-based exercises covering personalization, segmentation, and branching logic.
Course Customization Options
- To request customized training for this course, please contact us to arrange your specific needs.
Le Chat Enterprise: Private ChatOps, Integrations & Admin Controls
14 HoursLe Chat Enterprise is a private ChatOps solution that provides secure, customizable, and governed conversational AI capabilities for organizations, with support for RBAC, SSO, connectors, and enterprise app integrations.
This instructor-led, live training (online or onsite) is aimed at intermediate-level product managers, IT leads, solution engineers, and security/compliance teams who wish to deploy, configure, and govern Le Chat Enterprise in enterprise environments.
By the end of this training, participants will be able to:
- Set up and configure Le Chat Enterprise for secure deployments.
- Enable RBAC, SSO, and compliance-driven controls.
- Integrate Le Chat with enterprise applications and data stores.
- Design and implement governance and admin playbooks for ChatOps.
Format of the Course
- Interactive lecture and discussion.
- Lots of exercises and practice.
- Hands-on implementation in a live-lab environment.
Course Customization Options
- To request a customized training for this course, please contact us to arrange.
Cost-Effective LLM Architectures: Mistral at Scale (Performance / Cost Engineering)
14 HoursMistral is a high-performance family of large language models optimized for cost-effective production deployment at scale.
This instructor-led, live training (online or onsite) is aimed at advanced-level infrastructure engineers, cloud architects, and MLOps leads who wish to design, deploy, and optimize Mistral-based architectures for maximum throughput and minimum cost.
By the end of this training, participants will be able to:
- Implement scalable deployment patterns for Mistral Medium 3.
- Apply batching, quantization, and efficient serving strategies.
- Optimize inference costs while maintaining performance.
- Design production-ready serving topologies for enterprise workloads.
Course Format
- Interactive lecture and discussion.
- Ample exercises and practice.
- Hands-on implementation in a live-lab environment.
Customization Options
- To request a customized training for this course, please contact us to arrange.
Productizing Conversational Assistants with Mistral Connectors & Integrations
14 HoursMistral AI serves as an open-source AI platform, empowering teams to construct and embed conversational assistants within both enterprise operations and customer-facing processes.
This instructor-led training, available online or onsite, is tailored for beginner to intermediate-level product managers, full-stack developers, and integration engineers looking to design, integrate, and productize conversational assistants using Mistral's connectors and integrations.
Upon completion of this training, participants will be capable of:
- Connecting Mistral conversational models with enterprise and SaaS connectors.
- Implementing retrieval-augmented generation (RAG) to ensure grounded responses.
- Creating UX patterns for both internal and external chat assistants.
- Deploying assistants into product workflows to address real-world use cases.
Course Format
- Interactive lectures and discussions.
- Practical integration exercises.
- Live laboratory sessions for developing conversational assistants.
Course Customization Options
- For customized training requirements, please contact us to make arrangements.
Enterprise-Grade Deployments with Mistral Medium 3
14 HoursMistral Medium 3 is a high-performance, multimodal large language model engineered for robust, production-grade deployment within enterprise settings.
This instructor-led training, available either online or on-site, is tailored for AI/ML engineers, platform architects, and MLOps teams at intermediate to advanced levels. The course focuses on deploying, optimizing, and securing Mistral Medium 3 for specific enterprise applications.
Upon completion of this training, participants will be equipped to:
- Deploy Mistral Medium 3 via API or self-hosted solutions.
- Enhance inference performance while managing costs.
- Develop and implement multimodal use cases using Mistral Medium 3.
- Apply security and compliance best practices suitable for enterprise environments.
Course Format
- Interactive lectures and group discussions.
- Extensive exercises and practical sessions.
- Hands-on implementation within a live-lab environment.
Course Customization Options
- For customized training arrangements, please contact us directly.
Mistral for Responsible AI: Privacy, Data Residency & Enterprise Controls
14 HoursMistral AI is an open and enterprise-ready AI platform that provides features for secure, compliant, and responsible AI deployment.
This instructor-led, live training (online or onsite) is aimed at intermediate-level compliance leads, security architects, and legal/ops stakeholders who wish to implement responsible AI practices with Mistral by leveraging privacy, data residency, and enterprise control mechanisms.
By the end of this training, participants will be able to:
- Implement privacy-preserving techniques in Mistral deployments.
- Apply data residency strategies to meet regulatory requirements.
- Set up enterprise-grade controls such as RBAC, SSO, and audit logs.
- Evaluate vendor and deployment options for compliance alignment.
Format of the Course
- Interactive lecture and discussion.
- Compliance-focused case studies and exercises.
- Hands-on implementation of enterprise AI controls.
Course Customization Options
- To request a customized training for this course, please contact us to arrange.
Multimodal Applications with Mistral Models (Vision, OCR, & Document Understanding)
14 HoursMistral models are open-source AI technologies that have expanded into multimodal workflows, supporting both language and vision tasks for enterprise and research applications.
This instructor-led, live training (available online or onsite) targets intermediate-level ML researchers, applied engineers, and product teams looking to build multimodal applications using Mistral models, including OCR and document comprehension pipelines.
By the end of this training, participants will be able to:
- Set up and configure Mistral models for multimodal tasks.
- Implement OCR workflows and integrate them with NLP pipelines.
- Design document comprehension applications for enterprise use cases.
- Develop vision-text search and assistive UI functionalities.
Format of the Course
- Interactive lecture and discussion.
- Hands-on coding exercises.
- Live-lab implementation of multimodal pipelines.
Course Customization Options
- To request a customized training for this course, please contact us to arrange.
