Ascendra

Talent is everywhere, but opportunity is not. Ascendra helps students break through unfair career barriers with fair benchmarking, structured practice, and merit-based progression.

Comment

https://www.loom.com/share/e02b3aaff4fa4c07b56a34931f9d6184 -- this is the video link; the video required a youtube link when we recorded on loom and the loom video is paywalled

Inspiration

Ascendra was inspired by a simple problem: access to career insight is deeply unequal.

Students from stronger socioeconomic backgrounds often benefit from advantages that are hard to see from the outside: family networks, insider knowledge, polished guidance on applications, better examples of successful CVs, awareness of what firms are actually looking for, and a clearer understanding of how to prepare for interviews and assessments. A huge amount of useful career data exists, but it is fragmented, hidden behind networks, or simply much easier to access if you already have the right environment around you.

We wanted to build something that helps democratise that information.

Our goal with Ascendra was to take data that is usually exclusive, scattered, or inaccessible, and make it available through a platform that is:

  • structured
  • interactive
  • motivating
  • genuinely useful

Rather than treating career prep as a black box, we wanted to turn it into something users can explore, understand, and improve at systematically. That is where both the social mobility mission and the Creative Data Manipulation theme came together for us.

What it does

Ascendra is a gamified career-prep and applicant benchmarking platform for students and early-career candidates.

Users choose a pathway such as:

  • software engineering
  • quant
  • investment banking
  • consulting
  • product
  • law
  • general graduate schemes

They also choose the application cycle they are targeting, such as:

  • spring weeks
  • internships
  • placement years
  • graduate roles

From there, Ascendra benchmarks them against realistic peer groups and helps them improve through three arenas:

  • Profile Arena for CVs, cover letters, and written answers
  • Professional Arena for behavioural, HireVue-style, and communication-based responses
  • Technical Arena for coding, logic, quantitative, and technical drills

The platform gives users:

  • Profile Elo
  • Professional Elo
  • Technical Elo
  • a weighted Overall Domain Elo
  • cohort-based benchmarking
  • a skill tree with repair paths
  • quests and boss battles
  • replay pages with judge feedback
  • market-fit recommendations based on external hiring data

What makes Ascendra distinctive is how it uses data.

Instead of just storing user inputs, we built a system that:

  1. ingests external/public hiring data
  2. converts it into structured signals
  3. compares those signals against the user profile
  4. feeds the results back into quests, challenges, boss battles, and skill-tree recommendations

This is where the Creative Data Manipulation theme is central. We are not just visualising data or building dashboards. We are transforming messy, unequal, hard-to-access career information into a training system that ordinary users can benefit from.

A key part of this is the mathematical engine underneath the platform.

For ranked updates, we use a confidence-aware Elo-style update:

[ \Delta E = K \cdot C \cdot (1-D) \cdot (S-\hat S) ]

where:

  • (K) is the base update factor
  • (C) is the aggregated confidence from the judge panel
  • (D) is disagreement between judges
  • (S) is the actual result
  • (\hat S) is the expected result

The expected result is:

[ \hat S = \frac{1}{1 + 10^{-(E_u - E_o)/400}} ]

We also calculate a weighted overall rating:

[ E_{\text{overall}} = w_P E_P + w_{Pr} E_{Pr} + w_T E_T ]

This lets different pathways weight different skills appropriately.

We also built a market-fit score:

[ F = 100\left(\alpha \cdot \cos(u,m) + \beta \cdot \text{Coverage}(u,m) + \gamma E - \delta G\right) ]

This compares a user signal vector (u) against a market demand vector (m), combining similarity, coverage, evidence quality, and critical gaps.

We also introduced custom answer-quality metrics such as:

[ \text{Signal Density} = \frac{\text{meaningful evidence units}}{\text{word count}} ]

Together, these systems help turn vague career preparation into something measurable, transparent, and actionable.

How we built it

We built Ascendra as a full-stack application with three connected layers:

  1. the product layer
  2. the mathematical scoring layer
  3. the market-intelligence data layer

Frontend

We built the frontend as a retro-inspired, interactive platform designed to make improvement feel visible and motivating, while still treating career preparation seriously.

It includes:

  • a dark-mode, pixel-styled interface
  • arena-based progression
  • pixel-art AI judges
  • skill-tree navigation
  • retro-styled Elo ladders and rank badges
  • replay pages that feel like after-action reviews

We wanted the experience to feel fun and engaging without losing credibility. The platform should feel approachable to a student who may not already have insider guidance, while still being rigorous enough to provide meaningful feedback.

Backend

The backend was built around structured domain entities such as:

  • users and domain profiles
  • documents and uploads
  • challenge templates and attempts
  • matches and scoring results
  • skill nodes and mastery
  • quests and benchmark packs
  • leaderboard and rating snapshots

On top of that, we added a market-data layer with models for:

  • raw external job postings
  • normalized job-demand profiles
  • market-fit reports
  • interview question entries
  • benchmark statistics

Creative Data Manipulation

This was the most important part of the build.

We focused on taking data that is normally messy, fragmented, or socially uneven in who can access it, and converting it into structured signals that could power the product.

The pipeline works like this:

  • ingest external/public hiring data
  • normalize it into skills, competencies, pathway hints, and assessment-type hints
  • compare those signals to user data extracted from CVs and challenge performance
  • translate the result into recommendations, quests, boss battles, and skill-tree suggestions

This is the heart of our interpretation of Creative Data Manipulation.

Instead of letting valuable hiring information remain exclusive to people with better networks, better schools, or more insider access, Ascendra tries to open that data up, structure it, and turn it into something everyone can use.

Scoring engine

We also built a multi-judge aggregation layer. Each judge produces:

  • a score
  • confidence
  • category-level outputs

The final score is:

[ S_{\text{final}} = \frac{\sum_i c_i s_i}{\sum_i c_i} ]

and disagreement is measured by:

[ D = \sqrt{\frac{\sum_i c_i (s_i - S_{\text{final}})^2}{\sum_i c_i}} ]

This helps make the system more trustworthy. If judges disagree, the outcome is flagged as lower confidence and Elo movement is reduced.

We also added cohort-normalized ranking so users are not just judged globally, but relative to realistic peers.

Challenges we ran into

One of the biggest challenges was designing a platform that felt fun and interactive, but still serious enough to support a mission around social mobility.

If the product was too game-like, it risked feeling trivial. If it was too rigid or corporate, it would lose the sense of momentum and accessibility that makes users want to keep improving. Finding that balance was difficult.

Another challenge was figuring out how to use data in a way that was genuinely creative and meaningful. It is easy to say “we use data”, but much harder to build a system that takes messy job descriptions, CV signals, challenge performance, and peer-group information, and turns them into something structured and actionable.

We also spent a lot of time thinking about fairness and trust:

  • how to avoid overconfident AI-style scoring
  • how to reflect judge disagreement
  • how to compare users fairly across cohorts
  • how to make sure we were not recreating the same prestige-driven inequalities we wanted to reduce

A major product challenge was making sure the platform helps users who may not already know the hidden rules of the process. That meant we had to focus on clarity, transparency, and actionable next steps rather than opaque scoring.

Accomplishments that we're proud of

We are most proud of the mission behind Ascendra and how it shaped the product.

1. Using data to support social mobility

We are proud that Ascendra is designed to help democratise access to career insight.

A lot of the data that matters in applications is not formally secret, but in practice it is unevenly distributed. Some students know what strong applicants look like, what firms value, how to present evidence well, and how to prepare for specific assessments. Many others do not.

We built Ascendra to help close that gap by turning fragmented hiring data into something structured, accessible, and useful for everyone.

2. Making Creative Data Manipulation central to the product

We are proud that data is not just decoration in Ascendra. It is the engine of the platform.

We take messy external hiring information and transform it into:

  • structured market signals
  • market-fit reports
  • skill-tree recommendations
  • targeted quests
  • benchmark packs
  • boss battles

That felt like a genuinely distinctive use of the theme.

3. Building a real mathematical foundation

We are proud that the platform is not just “AI feedback”. It has:

  • confidence-aware Elo
  • weighted domain ratings
  • multi-judge aggregation
  • cohort normalization
  • market-fit scoring
  • signal-density metrics

This made the system feel much more rigorous, transparent, and trustworthy.

4. Creating an engaging interface around serious information

We are also proud that we turned a serious and often intimidating process into something interactive and motivating through:

  • pixel-art AI judges
  • arena-based progression
  • skill-tree design
  • Elo ladders
  • replay-style feedback pages

What we learned

We learned that one of the most powerful uses of data is not just prediction, but translation.

Raw hiring data on its own is not enough. Job posts, CVs, interview answers, and benchmarking signals only become valuable when they are translated into a form people can actually act on.

We also learned that data access is a social issue. Information asymmetry is a real part of why career preparation feels unequal. Students with better networks often have better access to examples, patterns, expectations, and hidden standards. That shaped our thinking throughout the project.

Another key lesson was that gamification works best when it is tied to real progress. A ladder, a quest, or a badge only matters if it reflects something meaningful. That is why we kept trying to tie every progression system back to actual evidence, actual weaknesses, and actual market demand.

Finally, we learned that trust is everything in a product like this. If users do not understand how their scores are formed, or if the system feels arbitrary, the whole experience breaks down. That pushed us toward transparent formulas, confidence measures, and cohort-based comparisons.

What's next for Ascendra

The next step for Ascendra is to go further on both mission and depth.

We want to keep expanding the amount of useful career data we can structure and democratise, and keep turning that into systems that help users improve.

Next steps include:

  • richer market-intelligence sources and trend analysis
  • deeper benchmark packs and boss battles by pathway
  • stronger interview-question clustering for behavioural and HireVue prep
  • more advanced skill-tree logic and repair paths
  • better cohort analytics and benchmarking
  • more recruiter-calibrated benchmark examples
  • stronger fairness and integrity tooling
  • live mentor or interviewer modes
  • future institution-facing and employer-facing extensions

Long term, our vision is for Ascendra to help answer three questions for any user, regardless of background:

  1. Where do I stand right now?
  2. What exactly is holding me back?
  3. What should I do next to improve?

Ultimately, we want Ascendra to help make career preparation less opaque, less exclusive, and less dependent on privilege by transforming hidden or fragmented career data into a platform that is open, interactive, and empowering.

Built With

Share this project:

Updates