Edo Sanjaya Perkasa

Content Refresh Opportunity Model

FlyRank AI Internship · Python, pandas, scikit-learn, NumPy · 2026

• Ranked 30,000 existing pages by likelihood of losing search visibility, for a content lead with more pages than they can open and no time to work through them by hand.
• Found the label could be reconstructed from the features — 3,388 pages with no prior impressions had a declining rate of exactly zero by construction.
• The model was separating pages that couldn't decline from pages that could. A filter, not a prediction.
• Rescored on the population where decline was arithmetically possible, withdrew the figure I had already published, and wrote the retraction into the report itself.
• Built the held-out split in NumPy and measured what it bought — 0.901 against a random split's 0.908 — rather than assuming the split was honest.
• Write-up, notebook, and named failure modes: the cases page.

Lead Generation and Scoring Agent

Claude API · 2026

• Built for businesses expanding into new markets, who need the right partners and real contacts, not a list to work through.
• Agent generates and scores leads against the criteria that matter for the specific expansion, so the output is ordered by fit rather than by whatever the search returned.
• Built under a zero-cost constraint: no paid data provider, no subscription tooling.

Menu and Customer Ranking

pandas, sentence-transformers, HDBSCAN · 2026

• Self-directed build, briefed by my own F&B operating experience and reviewed with a business colleague on what an owner actually needs to decide: menu, which customers to approach, and which ads bring buyers.
• Grouped customers from a 500-row Indonesian F&B dataset built and cleaned from scratch, using sentence-transformers embeddings and HDBSCAN clustering.
• Output was a ranked menu list and named customer categories — the form an owner could act on, rather than a cluster plot to interpret.

Machine Learning Engineering Intern

FlyRank AI — Remote · July 2026 – Present

• Weekly graded deliverables on warehouse data in Python, pandas, and Jupyter: embeddings and clustering, intent and opportunity modelling.
• Audit models for leakage before reporting results — checking whether a feature can reconstruct the label, and whether a score survives on the population where the question is actually answerable.
• Benchmark every model against a stated baseline, including hand-written rules, so a score is read as a gap over something rather than on its own.
• Document failure modes and their fixes alongside results, with provenance written into the scripts that generate reports rather than only into the reports.

Founder

Pinya Co., Bandung · June 2026 – Present

• Juice and beverage business, building toward premium event and wedding catering.
• Built a Notion-based financial reporting system linking daily entries to monthly P&L summaries.

AI Data Annotator & Model Evaluator

RWS Group, Remote · June 2026 – Present

• Rank model outputs by human preference and evaluate response quality against project rubrics.
• Apply detailed annotation guidelines consistently across Indonesian and Mandarin locales, to fixed deadlines with per-task quality thresholds.

Data Annotator

CrowdGen, Remote · June 2026 – Present

• Label and review text data against detailed task specifications, maintaining consistency across large task volumes.
• Exercise judgment on ambiguous and edge-case items where the guidelines are underspecified.

Founder & Business Owner

Selecta Bakery, West Java · 2013 – August 2026

• Built and ran the business to a team of 20, with a repeat-customer base sustained over thirteen years.
• Wrote and enforced SOPs across all 20 staff, cutting refund and replacement requests by ~85% against written complaint records.
• Ranked customers, products, and priorities for thirteen years on handwritten and printed paper. The models I build now do the same job on data I can check.

Associate's Degree, Tourism and Hospitality Management

Temasek International School, Bandung · 2006–2008 · Endorsed by Raffles Academy, Singapore

O-Level

2004–2006

Certifications

Google AI Professional Certificate (7 courses) · Google × Coursera · July 2026
PMI-CPMAI Introduction · PMI · April 2026
Product Management: An Introduction · IBM × Coursera · March 2026