- Name
- Muhamad Ihsan Shahrir
- Role
- Product Manager · AI
- Industry
- EdTech SaaS · 4+ years
- Location
- Kuala Lumpur, Malaysia · UTC+8 · Open to remote
- Languages
- Bahasa Melayu (native) · English (professional)
Character sheet
About me
Most product failures aren't build failures. They're thinking failures: the wrong problem, solved confidently. My job is to make sure we're solving the right one, and to prove it before we scale it.
Right now I do that with AI. I own the AI roadmap for an education platform with 2M+ students and lead three engineering squads alongside design, QA and data. The domain is education; the craft is general: discovery, sharp specs, evaluation, and launches measured by what actually moved.
I care about products that are right, not just impressive. A confident wrong answer costs more than a slow right one, in any product. So I run evaluations before debating opinions, check the production data myself, and I'm not above opening the codebase and shipping the fix.
How I play
How I work
- Move 01
Find the real problem
The first explanation is usually a symptom. Find the cause before the fix.
Proof
Our question generator flunked Bahasa Melayu. The cause wasn’t the prompt: the curriculum doc didn’t match what’s actually tested.
- Move 02
Evidence before opinion
Loud opinions aren’t data. Neither are unquestioned leaderboards.
Proof
Two model evals picked different winners. I reconciled them and shipped a primary + fallback, not a favourite.
- Move 03
Cut to what proves it
Ship the smallest version that answers the question. The rest waits.
Proof
Class Insight v1 was one teacher, one class. The classroom badge shipped in 6 days.
- Move 04
Trust is a feature
Users forgive slow. Not confidently wrong.
Proof
Class Insight’s risk score comes from code. The AI only explains it, and a validator blocks invented numbers.
- Move 05
Charge at the moment of need
Give away the habit. Charge for what people reach for mid-task.
Proof
Free dashboard, AI analysis on a monthly quota, paywall shown mid-task: 12% free-to-premium conversion.
- Move 06
Get close to the code
Specs grounded in real data and real code earn engineers’ trust.
Proof
I query production data, commit code, and prototype before asking for a sprint.
Attributes
Strengths
Skill tree
Skills
AI Product
- RAG & context grounding
- Prompt engineering
- Agentic / multi-tool orchestration
- LLM evaluation & model benchmarking
- Multimodal OCR
- AI guardrails & content safety
Product Management
- PRD & spec writing
- Product discovery
- Roadmapping & backlog
- Go-to-market & launch
- Freemium monetisation design
Data & Research
- SQL
- Metabase
- Microsoft Clarity
- Funnel & cohort analysis
- User interviews & usability testing
Tools
- Claude Code
- Codex
- Antigravity
- VS Code
- Figma
- Supabase
- Sequel Ace
- n8n
- Lovable
- Google AI Studio
Level ups
Experience
- LV 4
Oct 2025 – now
Product Manager · Pandai Education
Own the AI roadmap for Pandai Teacher. Lead 3 engineering squads with design, QA and data.
- LV 3
Sep 2022 – Oct 2025
Associate Product Manager · Pandai Education
13 generative AI tools, Tutor Dashboard, learning-platform UX revamp (+75% MAU).
- LV 2
Feb 2022 – Aug 2022
Product Management Intern · Pandai Education
Ran 130+ live classes a week; +60% tuition-ops efficiency.
- LV 1
2022
BBA (Hons.) Retail Management · UiTM Puncak Alam
Plus a Diploma in Geomatic Science (UiTM Arau, 2017).
Certifications: Product Discovery & Product Launches · Product School, 2024