What's the AI Career Assistant?

The AI Career Assistant is a chatbot on your AIVIA dashboard that gives career guidance based on your actual evaluation results. It’s not generic advice — it reads your reports and knows AIVIA’s full components list, so its suggestions are grounded in what you’ve actually demonstrated.

What it knows about you:

  • Your evaluation scores across all rubric dimensions
  • Your strengths, gaps, and per-question feedback
  • Which domains and components you’ve been evaluated on
  • AIVIA’s full taxonomy of components — so it can map what you’ve proven to what else is out there

What you can ask:

Anything career-related. There are no fixed prompts — you type what’s on your mind. Some examples of what candidates ask:

  • “Match my evals to a job.” The assistant looks at your verified strengths and maps them to roles and components where those strengths matter. Useful when you’re job-hunting and want to target roles where your proven skills are the best fit.
  • “Prepare me for an interview.” If you have an interview coming up, the assistant can walk through your report’s gaps and strengths and help you think about how to address them in conversation. It draws from your evaluation results to help you anticipate where interviewers might probe and how to articulate what you know.
  • “How can I improve?” The assistant identifies which components or dimensions to work on next based on where your scores have the most room to grow. It can suggest which evaluation to take next, or which areas to study before retaking one.

What it doesn’t do:

It doesn’t write your resume. It doesn’t apply to jobs for you. It doesn’t guarantee anything about hiring outcomes. It’s a thinking partner that happens to know your verified skill profile in detail.

Why this is different from a general career chatbot:

A generic career chatbot gives generic advice. The AI Career Assistant starts from evidence — your scored, analyzed evaluation results — and works outward. When it says “your tradeoff analysis is strong but your prioritization needs work,” that’s not a personality quiz result. It’s drawn from how you actually performed under evaluation conditions.



Part 2: Reference Posts

Detailed reference material for both candidates and hiring teams. Voice: neutral third person, structured for clarity.


Reference 1. Rubric & Scoring

This post explains how AIVIA evaluations are scored — from individual answer dimensions through to the final grade.

Rubric dimensions

Every evaluation scores answers on two types of rubric dimensions:

Technical dimensions (3 visible + 1 hidden):

Evaluations operate in one of two modes. Each mode has its own set of technical dimensions:

Mode Dimension 1 Dimension 2 Dimension 3 Dimension 4
Debug Root Cause Isolation Evidence Interpretation Impact Assessment Hidden (LLM-generated from component)
Design Tradeoff Analysis Gap Identification Preventive Thinking Hidden (LLM-generated from component)

Candidates see dimensions 1–3 listed in the scenario post under “What You’ll Be Evaluated On.” Dimension 4 is generated by the LLM based on the specific component and is not shown in advance.

General dimensions (fixed, all modes):

  • Clarity
  • Prioritization
  • Reasoning Quality

These apply to every evaluation regardless of mode.

How dimensions are selected per answer

The evaluator picks 3–6 of the most relevant dimensions per response from the full pool of 7 (4 technical + 3 general). Not every dimension is scored on every answer.

Scoring scale

Score Label Meaning
1 Insufficient Missing, off-topic, or no understanding
2 Developing Relevant but incomplete or contains errors
3 Competent Adequate for the expected background level
4 Proficient Clear, correct, good reasoning
5 Excellent Exceptional insight, depth, or mastery

Coached answer cap

If a candidate provides an answer only after the evaluator hinted at it in a previous turn, that dimension is capped at 3/5. A score of 4 or 5 requires the candidate to lead the discovery independently.

Weighting

AIVIA applies two layers of weighting to scores:

Layer 1 — Follow-up depth (within each question thread):

Response Weight Rationale
Initial answer 1.0× Could be memorized
Follow-up 1 answer 1.2× Probes actual understanding
Follow-up 2 answer 1.5× Deepest probe, hardest to fake

Layer 2 — Recency (across question threads):

Thread Weight Rationale
Thread 1 1.0× Baseline
Thread 2 1.2× Consistency check
Thread 3 1.5× Ceiling test

The highest-weighted single response is Thread 3 / Follow-up 2 — the deepest probe in the final thread. This design rewards candidates who improve under pressure and penalizes those who collapse under follow-up probing.

Final grade

The final grade is pattern-based, not average-based. It is determined by the distribution of technical dimension scores across all 9 evaluator outputs (3 threads × 3 responses each):

Grade Pattern Requirement
Beginner >40% of scores are 1–2 Regardless of occasional highs
Intermediate ≥60% of scores are 3+ Majority competent
Expert ≥60% of scores are 4–5, with zero 1s Majority proficient, no failures

Additional factors:

  • Improving trajectory across threads can push a grade up.
  • Collapsing under follow-up pressure can push a grade down.
  • Technical dimensions are weighted higher than general dimensions.
  • The grade applies to the specific component evaluated, not to general engineering ability.

Final score

A weighted numeric score (on a 1–5 scale) is also computed. This score is used internally for badge aggregation but is not displayed to candidates or hiring teams.

Minimum data requirement

A valid grade requires at least 6 scored responses out of 9. If technical issues cause too many responses to go unscored, the result is marked “Insufficient Data” rather than assigned an inaccurate grade.


Reference 2. Badge Aggregation

This post explains how individual evaluation grades are aggregated into badges at the component, subdomain, and domain levels.

What badges represent

Badges are the visual representation of aggregated grades across AIVIA’s taxonomy. They appear at three levels:

  • Component badge — aggregated from evaluations for a specific component
  • Subdomain badge — aggregated from component badges within a subdomain
  • Domain badge — aggregated from subdomain badges within a domain

Each badge displays one of three grades: Beginner, Intermediate, or Expert. If there is not enough data to compute a grade, the badge shows “Not rated” or “More evals needed.”

How component badges are computed

  1. All eligible evaluations for a component are collected. An evaluation is eligible if it is completed and was taken at the Intermediate expected background level.
  2. If multiple evaluations exist for the same scenario (topic), only the most recent one counts.
  3. The component grade is the mode (most frequent grade) across the latest eligible evaluations.
  4. If there is a tie between two grades, the lower grade wins. The system is conservative — ties break downward.

How subdomain badges are computed

  1. All component badges within the subdomain are collected.
  2. A subdomain badge requires at least 2 graded components to be computed. Below that threshold, the subdomain shows “Not rated” or “More evals needed.”
  3. The subdomain grade is the mode across graded component grades.
  4. Ties break downward.

How domain badges are computed

  1. All subdomain badges within the domain are collected.
  2. A domain badge requires a minimum number of graded subdomains — either 3 or the total number of subdomains in that domain, whichever is smaller.
  3. The domain grade is the mode across graded subdomain grades.
  4. Ties break downward.

Why ties break downward

AIVIA’s badge system is intentionally conservative. A tied result means the evidence is split — and in that case, the system reflects the lower bound rather than the upper. This ensures badges represent demonstrated, consistent ability rather than optimistic interpretation.

Eligibility filter

Only evaluations taken at the Intermediate expected background level count toward badge aggregation. Evaluations taken at other expected background levels still appear in the candidate’s dashboard and reports but do not contribute to badge grades.

Relationship to individual evaluation grades

An individual evaluation produces a grade (Beginner, Intermediate, or Expert) based on the rubric scoring described in the Rubric & Scoring post. Badge aggregation then combines these individual grades upward through the taxonomy. The two systems are connected but distinct:

  • Evaluation grade = how a candidate performed on one specific evaluation
  • Badge grade = how a candidate’s performance aggregates across evaluations at a given taxonomy level

What candidates control

Candidates control which evaluations are visible to employers via the On Resume toggle. However, badge aggregation uses all eligible evaluations, not just those toggled On Resume. This means badges reflect the full picture of a candidate’s evaluated performance, regardless of which individual reports they choose to display.