Data Analyst · Business Process Optimization · AI & Automation

Keshin Ryandy

Robotics & AI Engineering graduate who turns business processes into data, automation and working software. At PT Astra International Tbk and PT United Tractors Tbk I mapped processes, gathered requirements, built Power BI dashboards and shipped automation with Python, Excel VBA and Power Automate. I build AI, so I know where it breaks, and I test every system before people rely on it.

Portrait of Keshin Ryandy
Rp70Mevent cost savings, Astra KM Fest 2026
64%engagement increase from a Power BI trend dashboard
1st PlacePLN ICE 2023, national EV design competition
3.60GPA, B.Eng. Robotics & AI, Universitas Airlangga
Live apps

Things I built that you can open

Web apps running in production or as public demos. Demos use sample data; company data is never included.

Case studies

Selected work

Business process optimization · Astra International · 2026

Digital event operations that saved up to Rp70 million

AI & Digital Initiatives · Web app, Microsoft Lists, Power Automate
S

A 3-track company event (Astra KM Fest 2026) needed registration, check-in and voucher claims, and the company license did not include Power Apps.

T

Redesign event operations across departments into one digital process, with existing tools and a tight timeline.

A

Mapped the as-is process with the organizing committee, HR and IT, then built a mobile QR check-in web app for ~10 staff, a live attendance dashboard per track and a verified e-voucher flow with digital signature and recap. Duplicate scans are flagged, invalid codes rejected, and manual code entry is the fallback.

Up to Rp70 million saved. 389 of 421 registrants (92%) checked in digitally, with every track above its 70% target.
Check-in dashboard on a phone showing 389 present of 421 registered Check-out dashboard on a phone showing 313 checked out
Live check-in and check-out dashboards used by event leaders on the day.
Business process optimization · Astra International · 2026

From paper forms to live on-screen results: KM Fest Pulse

AI & Digital Initiatives · Web app, JavaScript, live results dashboard
S

Pre-tests and post-tests were done on paper, tallied by hand, and never shown during the event.

T

Digitalize the assessment for three tracks running at once and make learning impact visible instantly.

A

Gathered requirements from the learning team, designed an anchored 1–5 self-assessment with a bilingual (ID/EN) mobile form, and built a results screen the admin shows on the projector. Load-tested before the event for hundreds of simultaneous submissions.

Paperless assessment with results live per track, e.g. 120 of 151 people responding in one session.
Audience in a ballroom answering a live poll shown on stage Pre-pulse results screen showing room average 3.0 out of 5
Pulse running live in Track 1, and the pre-session results screen.
Automation · Astra International · 2026

Automated personalized invitations and reminders

Python (Pillow, qrcode), Power Automate, Microsoft Lists, Outlook
S

Each of 422 participants needed an invitation with their name and a unique QR code, plus reminders. By hand: about 2 minutes each, roughly 14 hours in total.

T

Automate design generation and email delivery end to end.

A

A Python script renders each name and QR onto the event template; a Power Automate flow emails it, throttles delivery and logs “sent” per participant. AI helped map failure cases using dummy data only; real data stayed inside company systems.

95%+ less manual work: reminders for all 422 participants in one 10-minute run, each tracked by delivery status.
Flow
1

Microsoft Lists: participant data

2

Python: name + QR on template

3

Power Automate: email + reminders

4

Status log: sent, per participant

Automation · United Tractors · 2025

RPA bot that completes multi-branch forms automatically

Power Automate Desktop, Microsoft Excel, Google Forms
S

Training reports were typed into Google Forms by hand, and each program type needed a different form path.

T

Submit every Excel row automatically and choose the right branch for each program.

A

Built a flow with conditional branching, blank-field handling, an error-handling subflow and Success/Fail logging back to Excel.

Consistent submissions, fewer errors. A companion Excel-to-AI translation bot cut manual effort by 80%. Watch the demo
Power Automate Desktop flow for the form-filling bot
Data analytics · United Tractors · 2025

Power BI dashboard that turned search trends into content decisions

Power BI, Python (Pandas), Google Trends, Excel VBA
S

Marketing content was planned without data on what customers were searching for.

T

Deliver a data-driven view of customer interest to guide content strategy.

A

Collected and cleaned Google Trends data in Python, compared search volume with keyword difficulty, and built an interactive dashboard. VBA macros made data preparation one click.

64% increase in user engagement on UT Connect social media.
Power BI dashboard of search volume versus keyword difficulty
Data science · United Tractors · 2025

Customer segmentation with machine learning clustering

Python, Pandas, K-Means, DBSCAN, PCA
S

Behavioural data was large, noisy and unsegmented, so every customer was treated the same.

T

Find segments the business can act on.

A

Preprocessed and scaled the data, chose K with the elbow method, compared K-Means with DBSCAN (forced clusters vs isolated outliers), and reduced dimensions with PCA.

Explainable customer segments presented as heatmaps and 3D PCA plots.
3D PCA scatter plot of clusters
Document AI · Solo project

KTP-to-Excel data extraction

Python, YOLOv11, EasyOCR, Excel · synthetic dataset, no real IDs
S

Identity-card data was typed into Excel by hand: slow, error-prone and sensitive.

T

Turn KTP images into structured rows and find which extraction method is reliable.

A

Filtered blurry images with a sharpness threshold, labelled 18 field classes, trained YOLOv11 + EasyOCR and compared it with full-OCR keyword anchoring. Every row carries a confidence score; unreadable key fields are flagged “Needs Review”.

YOLO + OCR was more reliable on dense layouts; uncertain records go to a person instead of becoming confident-looking errors.
KTP extraction pipeline diagram
Rupiah banknote detected by the model
Computer vision · Project lead

Rupiah banknote detection

CNN in TensorFlow that recognizes Rupiah denominations and speaks the value aloud, so visually impaired users can identify notes on their own. Built from my own annotated dataset and tested on flipped and rotated notes.

GitHub →
NusaTinggi chatbot screen
NLP · Machine learning lead · Bangkit 2024

NusaTinggi hiking chatbot

Indonesian-language chatbot for beginner hikers. Built the NLP pipeline (tokenization, lemmatization) and an LSTM intent model, keeping answers tied to curated content because safety advice must be accurate.

GitHub →
Engineering & robotics

Hardware, control and design

Keshin with the winning electric motorbike and the PLN ICE 2023 first-place board
EV UNAIR · Electrical crew

1st Place, PLN ICE 2023

National electric-motorbike design competition. Drew the schematics, selected components and integrated the electrical system of “Saga”.

Team holding the CIWIL line-tracking robot
Robotics · Team leader

CIWIL line-tracking robot

Color Information Gatherer for Wired Line-tracking. Led all divisions, designed the mechanics and electronics, and programmed it on Arduino.

3D CAD model of a robot forklift
CAD · Team leader

Robot forklift in Autodesk Inventor

2D/3D design, technical drawings, stress analysis and animated mechanisms: a lifting arm and 360° mobility.

Industrial automation · Individual

PLC ladder logic for a mixing process

Omron CX-Programmer and CX-Designer: start/stop latch, level sensors, 100 s heater timer, cycle counter, interlocked outputs that never overlap, and an HMI for operators. Full cycle with automatic restart.

Watch the demo →

Also: a class schedule optimizer with Ant Colony Optimization, an A* pathfinding simulation in Pygame, and an undergraduate thesis on pothole volume measurement with SfM-MVS 3D reconstruction and Mask R-CNN.

Skills & experience

What I work with

Experience & education

PT Astra International Tbk · AI & Digital Initiatives, 2026
Event digitalization, AI use cases, workflow automation.

PT United Tractors Tbk · Data Science Intern, Operation Excellence, 2025
Power BI dashboards, RPA bots, ML clustering, process improvement.

Bangkit Academy 2024 · Machine Learning cohort
Industry program by Google, GoTo and Traveloka.

Universitas Airlangga · B.Eng. Robotics & AI Engineering, 2022–2026, GPA 3.60

Hard skills
Power BIExcel & VBAPython & PandasTensorFlowK-Means / DBSCANYOLO · Mask R-CNN · OCRNLP (LSTM)Power AutomateRPAAI BuilderMicrosoft ListsWeb scrapingJavaScript web appsPLC (CX-Programmer)HMI designAutoCADAutodesk InventorArduino
Certifications

TensorFlow Developer Certificate · Machine Learning Specialization (Stanford & DeepLearning.AI) · Data Analysis & Visualization with Power BI (Microsoft)

Contact

Let's build systems
people can trust.

Open to roles in data analytics, data science and AI, business process improvement and automation, including management trainee programs.