Jasper.ai

Inspiration

Interview preparation today is overwhelming and inefficient.

Candidates jump between platforms like LeetCode, blogs, and discussion forums with no clear structure or personalization. Everyone gets the same question lists regardless of their background, experience level, timeline, or target companies.

We asked ourselves:

What if interview prep adapted to you the way Gym trainer adapts to your goals and composition?

That question led to Personalized Interview Prep, an AI system designed to build a fully customized interview journey for every candidate based on their target company, their experiences and their timelines too!

What it does

Personalized Interview Prep builds a fully customized roadmap and actively tests your readiness.

Role-based Concept Prioritization

For a given role (e.g., Software Engineer, ML Engineer), the system:

Identifies all relevant concepts
Orders them by importance and interview frequency
Assigns impact-based weights

User Confidence Mapping

The user rates their confidence level for each concept.
This helps distinguish between:

High-priority weaknesses
Low-priority weaknesses
Strong areas

Timeline-Aware Roadmap Generation

Using:

Concept priority
User confidence
Available preparation time

We generate a tailored preparation roadmap that:

Focuses first on high-impact weak areas
Allocates time intelligently
Adjusts depth based on urgency
Avoids wasting time on already mastered topics

Role-Specific Mock Interviews + Actionable Feedback

The system conducts mock interviews tailored to the selected role:

Technical questions aligned with prioritized concepts
Behavioral questions relevant to the role
Difficulty adjusted to readiness level

After each mock interview, users receive:

Structured scoring
Concept-level performance breakdown
Specific, actionable improvement suggestions

This ensures users don’t just study, but also they validate their readiness.

How we built it

Priority Engine

We use LLMs to create a role-specific concept hierarchy that:

Maps interview frequency
Assigns importance weights
Differentiates core vs. secondary topics

Confidence Scoring System

Users self-assess their confidence per concept, which we:

Normalize into structured scores
Combine with concept priority
Use to compute a “focus score”

Roadmap Generator

Our roadmap engine:

Optimizes concept ordering
Distributes time proportionally
Accounts for preparation deadlines
Outputs a structured, time-aware plan

Mock Interview Engine

Generates role-specific interview questions
Evaluates answers using structured rubrics
Produces actionable, specific feedback

The system supports both planning and evaluation, giving users clarity on what to study and how they’re performing.

Challenges we ran into

Balancing Priority vs Confidence

A low-confidence topic isn’t always urgent.
We carefully balanced importance weight with skill gap size.

Making Feedback Truly Actionable

Generic AI feedback isn’t helpful.
We engineered structured rubrics to ensure specific and measurable suggestions.

URL Hallucination in Roadmap Resources

When generating learning resources within the roadmap, the model occasionally hallucinated URLs.
We addressed this by making the LLM validating links by scraping the web, and grounding recommendations in verified sources.

Feedback Latency

Generating detailed, rubric-based feedback introduced noticeable latency.
We optimized the evaluation pipelines to reduce response time while maintaining quality.

Accomplishments that we're proud of

Built a priority-weighted roadmap engine
Designed a scalable concept-ranking framework
Implemented role-specific mock interviews
Created structured, actionable feedback
Delivered a system that connects preparation with evaluation

Most importantly:

We transformed interview prep from passive studying into a focused, role-driven strategy with measurable validation. This not only helps us but all students who are out there preparing for interviews!