Desktop app · Early access open
Found 30 experiences
in 2,026 files
Uncover the experiences buried in your files, then finish a cover letter grounded in real company research, fast.
“I did a team project.”
One line becomes this story
From meeting notes, slides, and code, it finds your role, choices, setbacks, and numbers, with sources.
“I did a team project.”
- My role
- Owned seasonality-aware feature design and the model ensemble strategy.
GuidedProject_FinalDeck.pdflag_1 ~ lag_8 · rolling_mean_3, 6 · month_sin / cos
- Key decision
- Added an LSTM to LightGBM and XGBoost. The two tree models err in opposite directions, so they balance each other, and the LSTM learns long-range patterns that lag features can’t hold.
GuidedProject_FinalDeck.pdfBiases run in opposite directions, so the ensemble offsets them
- Setbacks
- Filled average prices logged as 0.0 with neighboring values. Also hit a limit: forecasting three periods ahead compounds the error.
GuidedProject_FinalDeck.pdfRecursive inference, so error accumulates further out (+3 periods)
- Results
- Baseline LightGBM: MAPE 3.2% for radish, 3.62% for dried chili.
GuidedProject_FinalDeck.pdfRadish MAPE: 3.2 · Dried chili MAPE: 3.62
Pick where to look
WeKnow scans your Desktop, Documents, Downloads, and iCloud Drive on its own. You can also pick folders yourself.
- Reads your GitHub, Google Drive, and Notion too
- Your originals stay put. Nothing is moved or changed
Your experiences land on one board
Scattered files are grouped by activity into one experience card. Instead of leaning on memory, WeKnow rebuilds each experience from the facts left in your files.
- Hover over a cover to see dates, a one-line summary, and skills
- Not just “academic club.” It pulls problem → your call → outcome as scenes from reports, decks, and meeting notes
- Even numbers that only lived in your files, like a 3% MAPE or 17,250 keywords, come with their source
Stop digging through folders
for one transcript
Awards, certificates, and language test scores issued in your name gather in one drawer. Copy the name, issuer, date, and score field by field to fit any application form. Language scores show their expiry date too.
You’ve found your experiences.
What does this company want to see?
It reads filings, press releases, and job postings, then picks experiences that fit what each question asks for.
- FilingNames better demand forecasting for fresh food a top priority
- PressRolls out AI ordering that adapts to farm-gate price swings
- PostingKey duties: building and running time-series forecasting models
- Do you frame a forecasting problem, like price or demand, with data?
- Do you explain why you chose that method?
- Did you back the result with numbers?
Farm-gate price swings, time-series forecasting, and a 3% MAPE hit all three points.
Fits the problem framing, but the forecasting angle is thin.
Sample screen · fictional company. Every line of company info cites its source, and anything not in the materials is marked as missing.
Your experience and company research,
visible in every sentence
See right on the draft which sentences came from your files and which from company research.
Pick a question,
get a draft with evidence
It picks episodes that fit the posting and the question, then writes a draft. Each sentence shows which file it came from.
- Sentences it couldn’t verify get a dotted underline and a quick “Right / Not quite” check
- Answer once, and it won’t ask again on your next application
Same experience, different tone Opening line only
With a baseline LightGBM model alone, I brought price error for radish and dried chili down to about 3%, then built an ensemble by adding XGBoost and an LSTM, which behave differently.
Answer first, evidence afterWhen weather and disasters rattle supply, crop prices swing with it. To read those swings ahead of time, the model first had to capture the past two months.
Set the scene, then the decision and what changedExperiences you didn’t have
get caught in the interview
So we built it to say “I don’t know” when it doesn’t.
- Assumptions are flagged
Anything your files don’t confirm gets a dotted underline, and you confirm it yourself. It never writes up things you didn’t do.
- Your files stay on your PC
Nothing is uploaded to a server. There’s also a fully local mode that sends nothing out.
- Unknowns stay blank
Company info comes only from filings and job postings. If it isn’t there, it says so.
We revise it together
until it’s where you want it
It flags vague sentences and swaps in specifics from your own experience.
Used various models to achieve high prediction performance.
A baseline LightGBM model reached a price error (MAPE) of 3.2% for radish and 3.62% for dried chili. I then built an ensemble by adding XGBoost, which errs in the other direction, and an LSTM that learns long-range patterns.
- Edit by chat
Say “add more numbers” or “cut the teamwork part,” and it revises just that. If your files fall short, WeKnow asks first.
- Reviewer
Reads your draft like a recruiter and flags where it misses the question or the evidence is thin.
- Proofreading
Catches filler like “various” with no numbers, stiff phrasing, and sentences where the subject and verb don’t agree. Numbers and proper nouns are left alone.
Try it free,
go Pro when you’re all in
All prices include VAT.
Free
KRW 0
Try it first
- 3 cover letter drafts
- Experience mapping (files up to 10MB)
- 5 AI chats a day
Pro
KRW 19,900 KRW 14,900 / mo
1-month pass · no auto-renewal
KRW 14,900/mo KRW 7,500/mo
6-month pass, KRW 45,000 · no auto-renewal
Sign up early, get your first month of Pro free
- Organize any folder on your PC
- Unlimited AI chat
- Unlimited cover letter drafts and edits
- In-depth company and role reports
- New features first
Pay once and use Pro for the period you choose. Nothing renews automatically. See our Payments & Refunds policy for refunds.
Questions?
Anything else? business@agentickor.com
When does it launch?
Once we set a date, you’ll be the first to hear about it at the email you signed up with.
Where does it run?
It’s a desktop app you download and install on your computer, because it reads your files right where they are.
Are my files uploaded to a server?
Your original files are never uploaded. Only the text needed for analysis goes to the AI model, after national ID numbers, bank accounts, and other people’s contact details are masked. There’s also a fully local mode that sends nothing.
What’s an episode?
One experience split into scenes, each big enough for a paragraph of a cover letter. Once made, episodes are reused for your next company’s cover letter.
Will I be charged for signing up early?
No. We only take your email, never payment details. When you start Pro after launch, your first month is free.
How do I cancel my sign-up?
Email business@agentickor.com from the address you signed up with, and we’ll delete it.