Sample. This is a made-up candidate, not a real person.
Jordan Avery and every answer on this page were written by the Aethel team to show what a finished interview looks like. The scores and quotes follow the same rules as real results.
A finished interview, from both sides
After the interview, the candidate sees their results first. If they took the interview through a company’s link, that company sees the same transcript and scores, and records its own decision.
What to look for
- Every score quotes the candidate’s own words, and you can find each quote in the transcript.
- Collaboration says “Not enough evidence” because the interview did not cover it. That is not a low score.
- The interviewer says it is an AI in its first line.
- Your team makes the hiring decision. The AI review does not recommend who to hire.
Jordan Avery's own results page.
Sample. This is a made-up candidate, not a real person.
Your interview results
Data and machine learning interview · Finished September 30, 2026 · 5 answers
Sample. This is a made-up candidate, not a real person.
Overall
74 / 100
Based on 3 of 4 areas. Areas without enough evidence are left out, not scored low.
What you explained
The candidate described an hourly pipeline that moves order events from Kafka through Spark into Snowflake. They explained fixing late events by reprocessing a six-hour window and merging on order ID, then checking the new job against the payments system for two weeks before switching. They described a one-page note with a chart that explained the change to finance. The interview did not ask about working with teammates, so collaboration was not scored. In your words: "I changed the job to reprocess a rolling six-hour window and merge on order ID, so a late event updates its row instead of being dropped or counted twice."
This is an AI review of the interview. Each score quotes your own words. People at a hiring company make any hiring decision.
Sample. This is a made-up candidate, not a real person.
Scores by area
Problem solving
78 / 100
In your words
“I changed the job to reprocess a rolling six-hour window and merge on order ID, so a late event updates its row instead of being dropped or counted twice.”
“I would add data quality checks from the start, like row counts and null checks on key columns, with an alert when one fails.”
Communication
70 / 100
In your words
“I wrote a one-page note that explained why past days would go up slightly, with a chart of old and new daily totals, and I presented it at their weekly review before we switched.”
Modeling and analysis
74 / 100
In your words
“I ran the old job and the new job side by side for two weeks and compared daily totals against the payments system, which is our source of truth.”
“I also added a test that feeds in out-of-order events and checks that the final table has exactly one row per order.”
Collaboration
Not scored
Not enough evidence in this interview to score this area.
Sample. This is a made-up candidate, not a real person.
Full transcript
11 messages, 5 answers. The questions came from an AI interviewer. Highlighted words are quoted as evidence for a score. The label shows which area.
AI interviewer
Hi, I’m Aethel’s AI interviewer, not a person. I’ll ask about your own work for about 15 minutes, and a written transcript is kept. To start, tell me about a data pipeline you built or own. What does it do?
You
I own the pipeline that loads order events into our warehouse. Events come from the checkout service through Kafka, a Spark job cleans them every hour, and the results land in Snowflake tables that the finance dashboards read. It handles about two million events a day.
AI interviewer
What was the hardest problem you ran into with that pipeline?
You
Late events. Some orders arrived hours after the hourly job had already run, so daily revenue in the dashboards came out about 3% low until someone reran the job by hand. I changed the job to reprocess a rolling six-hour window and merge on order ID, so a late event updates its row instead of being dropped or counted twice.Evidence for Problem solving After that, the daily totals stopped drifting.
AI interviewer
How did you check that the change was right before you switched over?
You
I ran the old job and the new job side by side for two weeks and compared daily totals against the payments system, which is our source of truth.Evidence for Modeling and analysis I also added a test that feeds in out-of-order events and checks that the final table has exactly one row per order.Evidence for Modeling and analysis The new job matched payments within a few dollars each day.
AI interviewer
Who relies on these numbers, and how did you tell them about the change?
You
Mostly people in finance. I wrote a one-page note that explained why past days would go up slightly, with a chart of old and new daily totals, and I presented it at their weekly review before we switched.Evidence for Communication
AI interviewer
If you rebuilt this pipeline today, what would you do differently?
You
I would add data quality checks from the start, like row counts and null checks on key columns, with an alert when one fails.Evidence for Problem solving We only found the late-event issue because someone in finance noticed the numbers looked off. I would rather catch it first.
AI interviewer
Thank you. That’s the end of the interview. Your results will be ready in a moment.
Sample. This is a made-up candidate, not a real person.
Also on a real results page
- Feedback on what the interview showed, and what it did not.
- A private share link the candidate can turn off at any time.
- A way to ask a person at Aethel to review the result.
- Buttons to download or delete their data.
Try it for real
Experts take the interview for free and see their own results. Companies send an interview link and review results like these.
Want the details first? Read how interviews work.