jay-0804/jaiib-synthetic-intent

JAIIB Synthetic Intent Dataset

A synthetic dataset of people who intend to attempt the JAIIB (Junior Associate of the Indian Institute of Bankers) certification, built to model the audience of an exam-prep provider. It captures who wants to take JAIIB, why, how ready they are, and what blocks their preparation.

⚠️ Synthetic / illustrative. Every value is generated from designed assumptions, not measured from real people. The structure (personas, eligibility rules) is grounded in real IIBF facts, but the distributions and correlations reflect modelling choices. Use it for pipeline development, prototyping and proof-of-concept — not as real market data or to drive business decisions.

How it was generated

Built with NVIDIA NeMo Data Designer (open-source library). Each record is assigned to one of seven personas; structured fields are produced by Data Designer samplers (Gaussian for numeric scores, category for categorical fields) run one batch per persona, and readiness_score / enrollment_likelihood are Data Designer expression columns derived from the others. No personal data is involved.

Records

3,000 rows, 24 columns, no missing values.

Personas

Persona	Share	Notes
Working Banker – Promotion Seeker	24%	Promotion/increment driven, time-pressed
Working Banker – Fresh Recruit	16%	Newly joined, building fundamentals
Working Banker – Near Retirement	8%	Long study gap, high anxiety
Bank-Exam Aspirant / Student	22%	Wants JAIIB but not yet eligible
Career Switcher into Banking	12%	Entering banking from another sector
Finance Professional (NBFC/fintech/coop)	10%	Adjacent finance roles
General Finance Learner	8%	Learning for its own sake

Schema (selected)

• persona_type, current_status, eligibility_status — segment and whether the person can currently sit JAIIB (only serving bank/FI employees are eligible).
• Demographics — age, gender, state, occupation, experience_years, monthly_income.
• jaiib_intent_score — strength of intent to attempt (1–10).
• Pain points (1–10) — time_constraint, workload, family_commitment, lack_of_guidance, exam_anxiety, confidence_gap, study_gap.
• readiness_score — derived: intent minus barriers (1–10).
• enrollment_likelihood — derived signal for course-purchase propensity (1–10).
• jaiib_awareness_score, primary_motivation, preferred_learning_mode, device_preference.

Files

• jaiib_dataset.csv — the dataset.
• charts/ — eight matplotlib analytics charts.

Making it real-world valid

Calibrate the persona shares and score means to a real survey or real user analytics, or use Data Designer's seed-dataset mode — the same pipeline then produces a grounded dataset with almost no code change.