Open menu

Valentino Allen Prasetyo

I build machine learning projects and ship the software around them.

Computer Science (AI) student at BINUS. I build ML projects end to end and have shipped two client websites.

View projectsDownload CV

AI Engineer & Data Scientist internships, from Feb 2027Jakarta / TangerangGraduating 2028

Reading Indonesian reviews one aspect at a time.

NLP course project, team of 2. Course paper, unpublished.

IndoBERT, fine-tuned to score five aspects in every review.

  1. Read the review
  2. Find the aspects
  3. Judge each one

“pengiriman sangat lama hampir semingguShipping tapi harganya murahPrice dan barangnya berkualitas bagusProduct quality”

  • Product qualityPositive98.7%
  • PricePositive98.4%
  • ShippingNegative96.2%
  • Seller serviceNot mentioned98.9%
  • PackagingNot mentioned98.3%

Walkthrough of real model output. IndoBERT, 7 Oct 2026. Highlights added by hand.

0.90

Recall on aspects a review never names. On named aspects: 0.91.

Sentiment macro-F1, test set

  • IndoBERT0.794
  • IndoBERTweet0.790
  • TF-IDF + LR0.537
  • TF-IDF + SVM0.345

BERT beats both baselines (McNemar, p < 0.001).

My role

Fine-tuning, evaluation and error analysis.

Weak spot

Only 9 neutral pairs and 10 price mentions in testing.

Built with

IndoBERT, IndoBERTweet, TF-IDF, Streamlit. 1,500 hand-labelled reviews.

Read the case studyWatch the demoCode on GitHub

Flagging dropout risk on the day a student enrolls.

Tabular ML, team of 3. Streamlit app and course paper.

Dropout risk from ten facts known at enrollment.

Recall
0.838
Share of real dropouts the app flags.
Precision
0.537
About 46% of flagged students did not drop out.
  1. Read the profile
  2. Estimate the risk
  3. Compare to the threshold

The app's default profile

Marital status
Single
Study program
Biofuel Production Technologies
Previous qualification
Secondary education
Mother's education
Secondary education
Father's education
Secondary education
Displaced student
No
Special needs
No
Gender
Female
Age at enrollment
18
International
No

34.0%Estimated dropout probability

Review threshold 40%

Below the threshold: not flagged.

Walkthrough of real model output. Random Forest, live app, 4 Oct 2026.

The research paper

Pre-acceptance data only. XGBoost did best, F1 0.667. I made the charts.

Bar chart of the ten most important XGBoost features in the Background-Application setting
Fig. 2. Top-10 XGBoost feature importance (Background-Application). Chart by Allen.
Distribution of predicted dropout probabilities for Graduate and Dropout students
Fig. 3. XGBoost prediction probability distribution. Chart by Allen.

My role

Trained the model and tuned its 0.40 threshold. Charts for the paper.

Weak spot

One public dataset. Every flag needs human review.

Built with

scikit-learn, XGBoost, Streamlit. UCI dataset, 3,630 students.

Read the case studyOpen the appRead the paper (PDF)

It may take a moment to wake up.

Measuring a foot from one photo and a bank card.

Computer vision course project, team of 5.

A photo has no scale. A card of known size fixes that.

My role

Frontend and part of the backend.

Weak spot

Accuracy is not validated yet. Dark or patterned floors cause failures.

Built with

Python, OpenCV and FastAPI.

  1. Find the card
  2. Correct the perspective
  3. Measure and classify
85.60 mmlength in mmwidth in mm

Width ÷ length gives the fit: narrow, normal or wide.

Illustration of the method. Classical OpenCV, no neural network.

Read the case studyOpen the demoCode on GitHub

Free hosting, so it takes about a minute to wake.

Shipped for clients

Nata Teknik catalog page: search, brand filters and a grid of hardware products with prices (447 listed)

Nata Teknik

Hardware retail · Bali · 2025

Product catalog and SEO storefront that replaced sharing the catalog by hand on WhatsApp.

“Allen built our site from scratch with real attention to detail. The structure is clean and our team can update content on our own without going back to a developer.”

Evan NataOperations Manager, Nata TeknikTranslated from Indonesian
natateknik.com
Pradipa Indonesia homepage: three leather cardholders on a wooden bench beside the headline “Handcrafted Elegance.”
Pradipa Indonesia landing page on a phone

Pradipa Indonesia

Leather goods · Surabaya · 2025

Brand landing page for a handmade leather wallet line.

“The results have been excellent. Allen stayed fast on responses throughout the project, the final site is very satisfying, and revisions were turned around quickly.”

Owen GabrielCEO, Pradipa IndonesiaTranslated from Indonesian
pradipaindonesia.com

About and contact

Computer Science (AI) at BINUS University, Tangerang. GPA 3.47 / 4.00, graduating in 2028.

valentinoallen0111@gmail.com

CV (PDF)