Diploma in Security Policies and Stability in RL
Online postgraduate diploma of 800 hours in policy safety and stability in RL: 8 modules covering risks in reinforcement learning, constrained RL, shields and safety filters, Lyapunov stability, policy robustness, formal verification of neural policies, and assurance and certification, with an assigned technical tutor.
Diploma in Security Policies and Stability in RL — studying from India
India's data and analytics capability is among the largest in the world, with major centres in Bengaluru, Hyderabad, Pune, Chennai and Gurugram. Global technology firms and domestic IT companies such as Tata Consultancy Services, Infosys, Wipro and HCLTech run large engineering and analytics operations. Engineering organisations increasingly apply AI, simulation and digital twin techniques to product design and manufacturing. Data protection rules are shaped by the Ministry of Electronics and Information Technology, while the National Association of Software and Service Companies represents the sector. Engineers here build models, analytics pipelines and simulation workflows across industries.
Fees for India: ₹137,000 (Indian Rupee), with instalments and employer sponsorship available.
Relevant bodies in India
Typical roles
Questions from India
Are there professional bodies in India that recognise data and AI engineering?
India does not have a statutory licence for data scientists or AI engineers. Relevant professional bodies include the Computer Society of India (CSI), the Institution of Engineers (India) and IEEE India sections. SEIUM's qualifications are private, not state degrees and not accredited by a national body, so their value depends on how employers view them.
Which employers in India hire data and AI professionals?
Demand is strong in IT services and consulting firms, banks and fintech companies, e-commerce and healthcare organisations, and research labs, especially in Bengaluru, Hyderabad, Pune and the NCR. A private online programme can support a career change or upskilling, but employers mainly assess coding, statistics and project portfolios alongside your qualifications.
Can I study this while working full time in India?
Yes. The programme is fully online and taught in English, so you can study from home without a student visa. Schedules are set in IST, so plan around your working hours. Specialist courses are about 300 hours, diplomas 800 and master's about 1,900. Each student has an assigned technical tutor, and fees can be paid in instalments. Ask your employer whether they will sponsor the fees.
Take the controls.
Run a real engineering trade-off from this field. Move the sliders, watch the model react in real time and beat the challenge.
What you will be able to do
Identify the specific risks of reinforcement learning in critical systems.
Formulate and solve RL problems with safety constraints.
Design shields and safety filters based on barriers and reachability.
Apply Lyapunov theory to certify policy stability.
Assess the robustness of agents against perturbations and attacks.
Use formal verification tools on neural policies.
Build safety arguments aligned with current regulations.
Who it's for
- AI safety engineers responsible for learned components in critical systems.
- Functional safety engineers who need to assess RL-based controllers.
- Safe RL researchers.
- Formal verification engineers extending their work to neural policies.
- Autonomous systems engineers.
- Critical software certification specialists.
- Engineering or science graduates aiming to enter the automotive sector.
What you'll work with
- IEC 61508:2010 (functional safety of E/E/PE safety-related systems)
- ISO 26262:2018 (road vehicles functional safety)
- ISO 21448:2022 (SOTIF) (safety of the intended functionality)
- ISO/PAS 8800:2024 (road vehicles, safety and artificial intelligence)
- ISO/IEC TR 5469:2024 (functional safety and AI systems)
- UL 4600 (safety of autonomous products)
- ISO/IEC 23894:2023 (AI risk management)
- Regulation (EU) 2024/1689 on Artificial Intelligence (EU AI Act)
- Safety-Gymnasium
- OmniSafe
- alpha-beta-CROWN
- PRISM
- Storm
8 modules
1.1Unsafe exploration and catastrophic states1.2Misspecification and reward hacking1.3Negative side effects1.4Distribution shift at deployment1.5Taxonomy of AI safety problems1.6Functional safety requirements and AI1.7STPA hazard analysis for agents1.8Risk acceptance criteria1.9Differences between safety and robustness1.10Workshop: STPA analysis of a control agent
2.1Constrained Markov decision processes2.2Lagrangian relaxation and primal-dual methods2.3Constrained Policy Optimization (CPO)2.4PPO-Lagrangian and TRPO-Lagrangian2.5Chance constraints and CVaR2.6Cost functions and safety thresholds2.7Safety during training and at deployment2.8Safety-Gymnasium test environments2.9Analysis of cumulative violations2.10Lab: PPO-Lagrangian in Safety-Gymnasium
3.1Shielding based on temporal logic3.2Control barrier functions3.3Safety filters via quadratic optimisation3.4Hamilton-Jacobi reachability analysis3.5Invariant sets and backup controllers3.6Simplex architecture for critical systems3.7Runtime supervisors3.8Interaction between shield and learning3.9The cost of conservatism3.10Practical exercise: CBF filter on a navigation agent
4.1Lyapunov stability in non-linear systems4.2Lyapunov functions learned with neural networks4.3Regions of attraction and their estimation4.4Input-to-state stability and robustness4.5Certificates with SOS programming4.6Contraction and incremental analysis4.7Policies with stability guarantees4.8Stability in hybrid systems4.9Robustness to delays and saturation4.10Lab: neural Lyapunov function for a pendulum
5.1Robust RL against adversarial perturbations5.2Attacks on observations and rewards5.3Adversarial training of agents5.4Robustness to dynamics variations5.5Domain randomisation and distributional robustness5.6Epistemic uncertainty with ensembles5.7Out-of-distribution state detection5.8Robustness and worst-case metrics5.9Scenario-based stress testing5.10Practical exercise: observation attacks on a PPO agent
6.1Network verification with Marabou and alpha-beta-CROWN6.2Bound propagation and abstract interpretation6.3Closed-loop verification6.4Probabilistic model checking with PRISM and Storm6.5Temporal logic specifications in LTL and STL6.6Simulation-based falsification6.7Monitoring STL specifications6.8Scalability and limits of verification6.9Integration into the development cycle6.10Lab: reachability verification of a controller
7.1Safety arguments for learned components7.2Safety cases with GSN notation7.3Requirements of ISO/PAS 8800 and ISO 214487.4ISO/IEC TR 5469 for functional safety and AI7.5Managing training data as evidence7.6Regression testing between policy versions7.7In-service monitoring and retraining7.8Responsibilities and traceability7.9Obligations under the EU AI Act7.10Case study: safety argument for a driving agent
8.1Case selection and hazard analysis8.2Constrained formulation8.3Training with safe methods8.4Safety filter design8.5Stability or reachability certificate8.6Robustness test campaign8.7Structured safety argument8.8Conclusions and residual risk8.9Report and presentation8.10Final project: final review with the tutor
How you'll learn
100% online training on the SEIUM campus, with downloadable material, recorded classes and an assigned technical tutor. There are no fixed timetables: you progress at your own pace within the intake, so it fits around your job.
Each module combines theory, guided reading of the original standard and applied exercises. The working software is Safety-Gymnasium, OmniSafe, alpha-beta-CROWN, PRISM and Storm, in free, academic or evaluation versions. Testing is studied on real procedures: constrained training in simulated safety environments, robustness testing against adversarial perturbations and closed-loop verification of neural controllers. Assessment is continuous by module, plus the final capstone project.
Standards covered: IEC 61508:2010, ISO 26262:2018, ISO 21448:2022 (SOTIF), ISO/PAS 8800:2024, ISO/IEC TR 5469:2024, UL 4600, ISO/IEC 23894:2023, Regulation (EU) 2024/1689 on Artificial Intelligence.
Build it end to end
Your final project follows the real engineering workflow, step by step, reviewed by your tutor.
Where this takes you
Requirements
- Prior technical training or professional experience in the sector. No university degree is required.
- An internet connection and a computer capable of running the programme software.
- English sufficient to read technical standards.
Ready to start?
Send your application today. Our admissions team replies within 72 hours with your study plan and funding options.
Frequently asked questions
The programme totals 800 hours over 6–8 months. You set your own pace within each intake, and your tutor helps you plan the workload around your job.
You work with Safety-Gymnasium, OmniSafe, alpha-beta-CROWN, PRISM and Storm, using free, academic or evaluation versions. Your tutor indicates the free alternative wherever one exists.
A SEIUM Institute diploma, awarded on passing the continuous assessment and the final project. It is SEIUM's own qualification, not an official state degree; its recognition depends on the programme completed and the regulations of each country.
Yes. Modules passed are recognised in a higher programme in the same area (postgraduate diploma or master's). Ask about the recognition policy when you enrol.
Prior technical training or professional experience in the sector (no university degree required), a computer able to run the programme software and English sufficient to read technical standards.
Start dates are set by SEIUM's faculty for each intake. Our admissions team will confirm the next available date when you apply.
Related programmes
Diploma in MILP/CP Modeling and Heuristics for Production
Online postgraduate diploma of 800 hours in MILP/CP modelling and heuristics for production. 8 modules covering production scheduling problems, mixed-integer linear programming, formulation improvement, constraint programming,…
Diploma in Scaling, Scheduling and Cost in HPC/Cloud
Online postgraduate diploma of 800 hours in scaling, scheduling and cost in HPC/cloud. 8 modules covering HPC and cloud infrastructures, application scalability, job scheduling with Slurm,…
Diploma in Geo-Temporal Predictive Models
Online postgraduate diploma of 800 hours in geo-temporal predictive models. 8 modules covering spatial and temporal dependence, classical statistical models, geospatial machine learning, spatio-temporal deep learning,…
More in Data, AI, Simulation & Analytics
All Data, AI, Simulation & Analytics programmes → · How to apply · Fees & financing · Scholarships