Frost & Sullivan MetaBrain Early Careers Program – AI Model Training & Evaluation Internship

Launch Your Career with Frost & Sullivan At Frost & Sullivan, we believe that the future belongs to curious minds, innovative thinkers, and problem-solvers who are eager to make an impact. We are inviting applications from postgraduate students, recent graduates, and early-career professionals with up to two years of experience to join our growing global teams across various business, technology, research, consulting, AI, data, and corporate functions. The Opportunity Frost & Sullivan is looking for intern roles supporting MetaBrain. The work combines applied AI, business understanding, structured knowledge, quantitative models and trustworthy engineering to transform research and advisory into reusable software-enabled services. Build more than a demonstration. Work with industry researchers, advisors and engineers to turn AI capability into tested decision-intelligence software that enterprises can use. Role Overview Show how your technical work supports a business problem. Research, consulting or advisory experience is preferred but not mandatory. Academic projects, thesis and reproducible research implementations are valid evidence; internships do not require prior full-time employment. Build measurable improvements to MetaBrain's research and advisory workflows. Translate domain-expert feedback into evaluation datasets, experiment designs and controlled changes to retrieval, model configuration or approved model training. Proposed engagement Stipend: Yes, paid internship. Duration: Preferably six months, possibility of an extension up to 12 months based on performance and where academic arrangements and work authorization permits are in place. Full-time Conversion: Depends on assessed performance, a suitable vacancy and eligibility; it is not guaranteed. Essential Requirements Master's/PhD study or qualification with substantial AI/ML training. Python and practical experience with at least one ML framework; knowledge of learning objectives, overfitting, dataset splits and evaluation. A project involving language models, natural-language processing, retrieval, model adaptation or rigorous ML evaluation. Ability to communicate results and uncertainty to non-technical colleagues. Preferred: Parameter-efficient adaptation, preference data, information retrieval, synthetic-data evaluation, annotation-quality measurement or experiment tracking. Business research, knowledge management or advisory exposure is desirable. For business-heavy work, strong analytical writing and numerical validation are especially valuable. Key responsibilities Create reliable evaluation assets: Work with advisors on annotation guidance, reference answers and domain-specific tests. Preserve evidence and provenance; separate training, validation and held-out data; manage versions and detect duplication or contamination. Run controlled experiments: Implement Python pipelines to compare model/prompt configurations, retrieval choices and domain-adaptation techniques. Use transparent baselines, ablation studies and repeatable settings rather than selecting only favorable examples. Connect metrics to business quality: Measure groundedness, numerical accuracy, relevance, appropriate abstention and task success. Calibrate automated grading against expert review; categorize errors and recommend improvements with cost and latency trade-offs. Deliver safe improvement cycles: Maintain experiment logs, regression tests, model/dataset documentation and rollback records. Support fine-tuning or preference-based learning only where data rights, expected benefits and platform approvals justify the work.