It finds what was never typed in
Notes, CVs, call transcripts and emails — not just the profile fields someone remembered to fill in.
AI assistant
Ask your database a question. Get an answer you can act on.
Kepler reads the notes, CVs, emails and call transcripts you can already open, answers with a record set you can save as a list, and turns the admin that follows into changes you confirm one at a time.
Who did we speak to about data engineering roles last quarter?
Five people, ranked by most recent contact. Sophia Bennett was yesterday, on the Northwind Capital intake call.
| Candidate | Current employer | Consultant | Last contact | Stage | Why it matched |
|---|---|---|---|---|---|
| Sophia Bennett | Monarch Analytics | Priya Nair | Yesterday | Interviewing | Intake call for the Northwind brief Call note · yesterday |
| Amir Hassan | Northwind Capital | Priya Nair | 2 Sep | Screening | Asked about Spark platform roles Email · 2 Sep |
| James Okafor | Juniper Systems | Tom Reilly | 29 Aug | Sourced | Screening call — data platform work Call note · 29 Aug |
| Mei Tan | Atlas Cloud | Tom Reilly | 21 Aug | Sourced | Replied about a London data role Email · 21 Aug |
| Elena Rossi | Forma | Lena Holm | 14 Aug | Sourced | Call note — analytics engineering Call note · 14 Aug |
Message Sophia on WhatsApp to confirm the interview on Thursday at 3pm
Here is the draft. Nothing is sent until you approve it.
Hi Sophia, confirming your interview with Northwind Capital on Thursday 10 September at 3pm. It is a 45-minute video call with their Head of Data. I will send the link tomorrow. Let me know if anything changes.
Sent from your connected WhatsApp number
Changes arrive as cards
When the answer is an edit, it comes back as a proposal with its source. Nothing is written until you confirm it.
It inherits your permissions
The assistant runs on your access. A record you cannot open is a record it cannot open either.
What you can ask
See it on your own data Five things it does, in the words you would use.
Search everything you hold, rank a pool against a job, count what happened last month, explain where a fact on a record came from, and draft the message that follows. Every answer comes back as something you can act on — a record set, a shortlist, a total, a draft.
Active data engineers in London who have used Kafka in production, available within a month
Found 4 candidates. Three mention Kafka on their CV; one said it on a recorded call last week.
| Candidate | Current employer | Notice | Why it matched |
|---|---|---|---|
| Sophia Bennett | Monarch Analytics | 1 month | Kafka streaming for card transactions at Halcyon CV |
| Amir Hassan | Northwind Capital | 2 weeks | Kafka feed for the risk platform CV |
| James Okafor | Juniper Systems | 3 months | Rebuilt the Kafka clickstream pipeline CV |
| Oliver Chen | Northwind Capital | Immediate | Mentioned running Kafka on the 2 Sep call Call note · 2 Sep |
Nothing lands on its own
When the answer is a change, it comes back as a card.
Move a stage, set a field, file a note, send a message — each one proposed with the source it came from, and applied only when you confirm it. Confirmed changes are written as you, in the record’s own history, where the rest of the team can see them.
Stage
ScreeningInterviewing
Northwind Capital asked for a first-stage interview on today’s call, 14 minutes in.
- Stage changed to Interviewing
by you · via the assistant · just now
- CV sent — Senior Data Engineer, Northwind Capital
by you · Tuesday
How the answers behave
It searches the words people used, and counts before it lists.
Three things that decide whether an answer is worth having: what it read, whether it gives you a number or a wall of rows, and what happens when the question is a morning’s work.
“led a team through a migration”6 matches
- Sophia Bennett“Managed four engineers across two squads through the warehouse migration.”CV
- Amir Hassan“Migrated on-prem Hadoop to EMR and S3.”CV
- Mei Tan“Hands-on until 2023; since then leading the platform team.”Call note · 21 Aug
It finds the phrasing you did not use
A search for “led a team through a migration” surfaces the CV that says exactly that — and the note and the call that say it differently.
- Priya Nair14
- Tom Reilly11
- Lena Holm8
Total33
It counts before it lists
Ask for a number and you get a number, with the breakdown behind it. The list only happens when you ask for one.
Re-score 214 candidatesbackground run
- Pull everyone against the specDone
- Score against the rubric133 of 214
- Draft the shortlist summaryQueued
Resumes where it stopped.
Long runs stay yours to stop
A big ask — re-score everyone against this spec — runs in the background, step by step, and pauses the moment you want a look.
The data boundary
It reads what you can read. It writes nothing without you.
The assistant lives inside the permissions you already have. That is the whole trick, and the whole boundary.
- It reads what you can open. The records, notes, CVs, emails and call transcripts your permissions already allow. Nothing else, nothing further.
- It writes only through a card. Every change is a proposal you confirm, written as you, in the record’s own history.
- It never trains on your data. Your database answers your questions. It does not become anyone else’s.
Questions, answered.
What can the AI assistant actually do?
Two things. It answers questions about your database in plain English — across notes, CVs, call transcripts and emails, not just profile fields — and it does the admin those answers imply: moving a stage, setting a field, filing a note, creating a task. Answers cite their sources, and changes arrive as cards you confirm.
How do I find candidates I have spoken to before without relying on my memory?
Ask in plain English. Kepler searches everything you can already open, including CVs, notes, call transcripts and emails, not only the fields someone filled in. A phrase like “led a team through a migration” surfaces the CV that says exactly that, and the ones that say it differently in a note or a transcript.
Can it change records without me?
No. Every edit comes back as a card — sourced, reversible, and applied only when you confirm it. Confirmed changes are written as you, in the record’s own history, not in a side channel nobody audits.
Can it see records I am not allowed to see?
No. The assistant inherits your permissions exactly: it reads what you can read and writes as you, never around you. Confidential records stay confidential, and your data is never used to train AI models.
What happens to a question that takes a while?
It runs in the background, step by step, so you can watch it progress and stop it whenever you want a look — it resumes where it stopped. Any set of results it produces can be saved as a list, so the work is still there tomorrow.
Start your free trial.
We set up your workspace, migrate your data and walk you through it.
- Free migration from your current system
- No card, no minimum seats
- A reply within one working day
Thanks, your request is in.
We will email you within one working day to set up your workspace and agree a time for the walkthrough.
Want a walkthrough sooner? Book a demo