GuroGabay: A MATATAG-aligned teaching copilot that gives Filipino teachers back their evenings

Problem Statement

DepEd’s teaching force of roughly 900,000 is rolling out the MATATAG curriculum, which means new competencies, new lesson formats, and new materials, often without enough textbooks. EDCOM 2 and teacher groups have repeatedly flagged heavy administrative and non-teaching workloads, and teachers commonly spend unpaid evenings and weekends writing lesson plans and reports. Teachers in multigrade and Indigenous Peoples schools, and in areas still using local languages under RA 12027’s provisions for monolingual classes, have almost no ready-made materials in their learners’ language. Burnout and uneven lesson quality follow, and learners end up with poorer outcomes, as shown in the Philippines’ low PISA 2022 results.

Proposed Solution

Guro AI is a web and mobile copilot built on an indexed library of the MATATAG curriculum guides and the most essential learning competencies. A teacher picks the grade, subject, quarter, and competency code. Guro AI then drafts a Daily Lesson Log or lesson exemplar in DepEd’s format, with activities localized to the community (for example, fractions taught through bilao-sized kakanin in Batangas), differentiated tasks for struggling and advanced learners, and a formative quiz with an answer key mapped to the competency. A single click translates or rewrites the material in Filipino, English, Cebuano, Hiligaynon, Ilocano, Waray, and more, with a “native speaker review” flag. Retrieval keeps every output tied to the official competency text. An admin assistant pre-fills school forms and reports from class records. Every output can be exported to DOCX, PDF, or Google Docs for printing in schools without internet access.

Target Market

• Primary users: public school teachers (about 900,000) and private school teachers, especially in K–10 and multigrade schools.
• Secondary users: master teachers and school heads who review lesson plans, and ALS mobile teachers.
• B2G buyers: DepEd central and regional offices, Schools Division Offices, and LGU Special Education Fund programs. B2B: private school networks and publishers.

How AI Is Used

• LLM with RAG over MATATAG curriculum guides: generates lesson plans that are aligned to competencies, not generic. A template library can’t localize or differentiate on demand.
• Multilingual generation and translation: produces materials in local languages, which barely exist commercially.
• Assessment item generation: creates quizzes with answer keys and plausible wrong options at different difficulty levels, tagged to competency codes.
• Document understanding (OCR plus LLM): reads a photo of an old lesson plan or textbook page and modernizes it to MATATAG.
• Alignment checker: a second LLM pass flags activities that don’t match the targeted competency, keeping quality high.

PLDT / Smart / ePLDT / Rokid Angle

• PLDT Enterprise and Smart: bundle the app with school connectivity programs and DICT or DepEd connectivity projects. Offer Smart education data promos for teachers.
• ePLDT cloud: host the curriculum index and teacher data in Philippine data centers. Sell it as a DepEd-wide SaaS, a strong B2G anchor.
• Offline export plus sync: downloads materials when the teacher is connected and syncs later, which suits low-bandwidth schools.
• Rokid (stretch): AR glasses that let a teacher see lesson cues hands-free during a science demo.

Demo Scenario (3 minutes for judges)

1. “Teacher Marites” handles a multigrade Grade 3–4 class in a mountain barangay in Iloilo. It’s 9 p.m. and she still has three lesson logs to write.
2. Live, she picks Grade 4 Math, Q2, the competency on equivalent fractions, and chooses Hiligaynon plus Filipino.
3. In about 30 seconds, Guro AI produces a complete Daily Lesson Log with a local hook (sharing a piece of piaya), differentiated tasks, and a 10-item quiz with an answer key.
4. The alignment checker flags one activity as off-competency, and she regenerates it with one tap.
5. She snaps a photo of an old worksheet, and Guro AI converts it into the MATATAG format.
6. Export to DOCX, ready to print at school tomorrow.
7. A counter shows “3 hours saved this week.” Close: “Para sa guro, hindi para palitan ang guro.” (For teachers, not to replace them.)

30-Hour Build Plan

Suggested stack: Next.js PWA, FastAPI, Gemini or GPT-class LLM, multilingual-e5 embeddings with Chroma, Google Cloud Vision or Tesseract OCR, docx and pdf generation libraries. Public data / APIs: DepEd MATATAG curriculum guides and the most essential learning competencies (public on the DepEd website), DepEd Commons and the Learning Resource Portal, the DepEd school directory on data.gov.ph, and Hugging Face Filipino and Philippine language corpora for evaluation.

Risks & Privacy Notes

• Hallucinated or culturally off content: retrieval grounding, the alignment checker, a mandatory teacher review step, and a “native speaker review” flag for low-resource languages.
• Weak local-language quality: start with major languages, let teachers rate and correct outputs, and use the corrections as a glossary.
• Data Privacy Act 2012: class records include minors’ data. Pseudonymize learner names before any LLM call, store data in PH, apply role-based access, and never use learner data for training without consent.
• Clause 24 compliance: the demo uses only public curriculum documents and synthetic class lists. AI-generated components are clearly disclosed.

AI Use Disclosure (draft)

We use a Gemini or GPT-class LLM with retrieval over public DepEd curriculum documents to generate lesson plans, quizzes, and translations, and an OCR model to digitize worksheets. Coding assistants helped write boilerplate, and an LLM helped edit this proposal. All learner data in the demo is synthetic. Team members and volunteer teachers reviewed sample outputs for accuracy and alignment.

Rubric Scoring