AI Intern – MCO Dimensioning

Nabízím práci
Jméno/Firma
Mavenir s.r.o.
Pracoviště
Holandská 868/5, Brno
Úvazek
Zkrácený
Požadované vzdělání
Bez maturity
Požadované jazyky
Angličtina
Profese
Informatika a IT služby
Nabízím práci/ Hledám práci
Nabízím práci
Vytvořeno
18. 11. 2026

O pozici

Pracovní nabídka

Role Summary
Join Mavenir's MVAS R&D team and explore how data-driven modelling and AI can improve the dimensioning of MCO, our modular telecom messaging platform. The aim is to recommend resource capacity, pod replica counts and u-service sizes for deployment configurations that have not been directly measured, while making uncertainty explicit.

This is a supervised, exploratory internship, initially focused on a selected subset of u-services and one end-to-end flow. You will work with experienced MCO development and test engineers, with access to the performance lab and domain experts, weekly syncs and milestone reviews. A clear negative result is a valid outcome; continuing cooperation may be considered if results are promising.

Part-time or full-time internship for 6–12 months.

Key Responsibilities
• Learn the MCO architecture and traffic flows; collect performance-lab measurements, deployment utilization data, KPIs and previous sizing decisions, and document their quality.

• Build CPU, memory, disk and network utilization models as functions of traffic profile for a selected subset of u-services, and combine them into a queueing-network model of a selected end-to-end flow.

• Estimate pod replica counts and u-service sizes (S/M/L/XL), including trade-offs between horizontal and vertical scaling; calibrate models from measurements and assess predictions for previously unmeasured configurations.

• Explore where AI helps with model fitting, missing-data handling and model composition, and where classical analytical or simulation methods are more appropriate.

• Handle missing inputs through documented assumptions, typical ranges and similar deployments; quantify uncertainty and sensitivity, report sizing ranges with confidence, and identify which customer inputs most need clarification.

• Prototype an AI-assisted questionnaire for sales teams to clarify missing information and map deployment needs to the required u-services.

• Recommend additional performance tests that would improve the model most effectively.

• Compare model accuracy and sizing margins with current practice; document findings, including negative results, and present recommendations on whether and how to continue.

Job Requirements

• Student or recent graduate in computer science, software engineering, telecommunications, applied mathematics/statistics or data science.

• Solid Python programming skills, including data analysis with tools such as pandas, NumPy and scikit-learn or equivalents.

• Basic knowledge of statistics and modelling: regression, uncertainty, confidence intervals and sensitivity analysis.

• University-level knowledge of queueing theory: M/M/1, M/M/c, queueing networks and multi-server capacity analysis.

• Basic understanding of computer networks and Linux, including the command line, log analysis and simple scripting.

• Practical experience with LLMs, including prompting, API usage and simple agents/RAG; study or personal projects are sufficient.

• Ability to work with incomplete and noisy data, critically assess results and communicate assumptions and limitations honestly.

• A structured and independent working style, clear communication and willingness to ask questions and accept guidance.

• Curiosity, reliability and a pragmatic approach: deliver in small iterations, prefer simple useful models and be prepared to report a negative result.

Good to Have (Nice to Have)
Relevant experience in any of the following is welcome; you do not need all of these skills.

• Telecom systems and protocols, such as SMS, MMS, SMPP, SIP, Diameter or mobile core networks.

• Capacity planning, performance testing or simulation tools, including discrete-event simulation.

• Monte Carlo or Bayesian methods for estimation with missing data or uncertainty.

• Time-series analysis or traffic forecasting.

• Containers/Kubernetes, monitoring tools such as Prometheus and Grafana, and Git.

Expected outcome
A tool or methodology that (semi-)automatically assists with dimensioning the MCO system for customer-specific deployment scenarios.

Success Criteria
The activity will be considered successful if it demonstrates one or more of the following:

· Meaningful prediction capability for previously unmeasured deployment scenarios.

· Reduction of manual effort required for sizing and dimensioning activities.

· Identification of a practical methodology that can be extended to additional MCO u-services and future customer deployment scenarios.

· Clear evidence that AI-assisted approaches provide value beyond existing dimensioning methods.

A well-documented negative result is also considered a successful outcome if it clearly demonstrates why the proposed approach is not sufficiently accurate, scalable, or practical for production use.