How Rithavo understands your career

Give Career Intelligence its own relational data model, separate from the single-JSON-blob Career Profile storage — versioning, provenance, and a curated Industry/Role taxonomy.

• 20 new intelligence tables (Evidence, Capability, Fit, Transition, Gap, Recommendation, Readiness, Diagnostic, Resume, Career Direction, and more)
• Versioning primitive (version / status / supersedes_id) reused by every later sprint
• ProvenancedField extended with evidence_ids and confidence
• Small, hand-curated Industry/Role reference taxonomy

Extract Evidence mechanically from what a person already entered — no AI, no invention, every excerpt verbatim and verifiable against the source.

• Rule-based Evidence extraction from CareerProfile fields
• Deduplication so re-running extraction never creates duplicate rows
• Anti-invention validation — every excerpt checked against the real source text

Answer 'what can this person demonstrate?' without ever treating a claim as proof.

• Four-tier capability classification: Claimed / Insufficient Evidence / Supported / Strongly Demonstrated
• Holistic, evidence-based confidence scoring (not a raw evidence-count formula)
• Capability clustering (Leadership, Analytical, Commercial, Technical, and more)

Separate three genuinely different questions instead of one fit score: can you do this today, how much would transfer, and how hard would the move be.

• Current Fit / Transferable Fit / Transition Effort as three distinct bands
• Confidence kept separate from fit — never invented, never a percentage
• Thin-taxonomy protection so incomplete data is never scored as a false good fit

Turn 'not a perfect match' into a specific, classified reason — not every unmet requirement is a real gap.

• Six-way gap classification (Skill / Experience / Evidence / Positioning / Information / No Meaningful Gap)
• Severity and priority modeling, so the system can name the 3 that matter, not list 15
• Conservative gap-dependency detection (only when the prerequisite is itself a real gap)

Don't sell the gap — help close it, with the lowest-cost path considered first.

• Gap-type-aware intervention taxonomy (Guided Learning, Practice, Document Evidence, Position Profile, Gather Information)
• Free-first defaults, with the paid alternative always shown, never assumed
• 'Completion of a course is not proof of capability' encoded directly in the data model

Explain readiness rather than score it — a synthesis layer, not a new judgment.

• Current / Transition / Target Readiness as three distinct, explainable dimensions
• Insufficient-information handling — missing data is never read as inability
• Confidence inherited from the underlying Fit assessment, never invented fresh

Let a resume diagnostic contribute evidence about the person, without ever confusing 'the resume communicates this poorly' with 'the person lacks this capability.'

• Diagnostic -> Finding -> Evidence pipeline with explicit RESUME_SPECIFIC / CAREER_LEVEL classification
• Resume versioning, diagnostic deduplication by run id, structural import-safety checks
• Explicit rubric-version compatibility boundary added post-review, so a future rubric change is rejected rather than silently misinterpreted

Answer 'which paths are worth exploring?' by scanning the taxonomy instead of requiring every candidate to be named in advance.

• Deterministic scan across the curated Role/Industry taxonomy, reusing the unmodified Fit engine
• Meaningful-fit threshold and thin-taxonomy exclusion — no invented ranking score
• Activated the dormant CareerDirectionAssessment entity, distinguishing the user's stated direction from Rithavo-discovered alternatives

Answer 'what changed since my last assessment, and why?' using only what was already computed and stored.

• Version-history walkers for CareerTransition, Readiness, and Career Direction (extending the pattern already proven for Capability)
• Deterministic version-to-version comparison for Capability, Gap, Readiness, and Career Direction
• Strict what-changed vs. why-it-changed separation — no causal claims invented

Answer 'given everything you know, what should I focus on next?' by composing the existing engines — never a new judgment engine.

• One coherent CareerIntelligenceSummary assembled from Fit, Gaps, Recommendations, Readiness, and Evolution — all read-only
• Recommendation-decision-aware filtering (a rejected recommendation never resurfaces as a fresh ask)
• Cross-engine consistency verified directly against the underlying persisted rows

Make eleven sprints of Career Intelligence visible, understandable, and useful through the actual website — the same structured data, presented, never recomputed.

• Career Intelligence overview, Career Paths, and Rithavo Build (this timeline) pages
• Provenance preserved visually throughout (USER_PROVIDED / DIAGNOSTIC / RITHAVO_DERIVED / RITHAVO_RECOMMENDED)
• Demo Mode using synthetic data only
• Homepage storytelling: hero, guided product tour, and human/aspirational journey, connecting straight into the live product — no separate frontend
• Rithavo Build reframed in plain language for non-technical visitors, real sprint history kept intact underneath
• Fit engine direct-experience bump made single-pass so the displayed band and its written reasoning can never disagree, with a permanent regression test