top of page
Insights
What we're Learning.
Field notes from production, not thought leadership. We publish when we have something to say.
All Posts
Credit Risk Modelling with AI: A Guide for Malaysian Lenders
How Malaysian banks and lenders can build AI credit risk models that meet BNM expectations: PD, LGD and EAD, explainability, validation and governance.
AI Runbooks: How to Run AI Agents in Production
AI runbooks explained for Malaysian enterprises: what to include, how they differ from Azure scripts, and how to meet BNM and PDPA expectations.
Experiment Tracking for ML: A Guide for Malaysian Enterprises
How to track ML experiments so results are reproducible and audit-ready: what to log, how to choose a tracker, and what Malaysian PDPA rules mean for you.
Feature Engineering Techniques for Production ML
Feature engineering techniques that hold up in production: encoding, scaling, time features, selection and automation, with Malaysian PDPA considerations.
ML Model Validation: A Practical Guide for Enterprise Teams
How to validate ML models for production: the process, techniques, verification vs validation, and what Malaysian regulators expect from audit-ready evidence.
Managed MLOps: What It Covers and How to Buy It
Managed MLOps puts deployment, monitoring and retraining of your models under one accountable team. See what it covers, what it costs you in control, and how to choose.
AI Business Case: How to Build One Malaysian Boards Approve
How to build an AI business case for a Malaysian enterprise: hard and soft benefits, costs, governance, BNM and AIGE guidance, and a board-ready structure.
Batch vs. Streaming: When Seconds Justify Continuous Processing
Compare batch, streaming, and hybrid pipelines by latency, throughput, cost, and operational load, then choose the simplest model that meets your SLA.
Enterprise Chatbot Architecture: Keep RAG Private, Tools Narrow
Build enterprise chatbots with private networks, tenant specific retrieval, narrow agent tools, and audit trails backed by Microsoft and OWASP guidance.
Model Risk Management: 6 Lifecycle Controls for Finance and AI Teams
Learn how finance and AI teams can manage model risk across its lifecycle, apply controls in proportion to impact, and put MAS guidance into practice.
AI Portfolio Management for Enterprises: One Portfolio, No Handoffs
Manage enterprise AI portfolios with NIST and Malaysia’s MOSTI AIGE guidance for intake, risk scoring, stage gates, and fewer handoffs by design.
AI Integration Patterns for Architects: RAG vs. Agents in Production
Compare five enterprise AI integration patterns, weigh RAG against agents, and see how architects manage governance, security, and production delivery.
RPA vs AI: Why Enterprise Workflows Need Human Checks
RPA and AI handle different work: bots execute fixed rules, while AI interprets ambiguity. Learn when to combine them and build in human review for...
AI Vendors Can Lock Enterprises In: Test Exit Costs in One Day
AI vendors can lock enterprises in. Audit exit costs and run a one day second source test, then use contract and engineering checks to prove portability.
Enterprise AI Roadmap for Leaders: From Pilots to Governed Production
Help enterprise leaders move AI beyond pilots with a phased roadmap for strategy, data, governance, and production, plus an applied delivery playbook.
6 Stage Pipeline for Production Bank Statement Extraction in Malaysia
Practical production advice for finance teams to build a modular, 6 stage bank statement extraction pipeline. Covers Bankstatemently benchmarks and...
AI Data Governance: ISO/IEC 42001 + Malaysia PDP Mapped for Enterprise
Enterprise playbook for AI data governance mapped to ISO/IEC 42001 and Malaysia's PDP consultation. Prioritised, field-tested steps and a checklist you...
Engineers: Make Retrieval Augmented Chatbots Reliable with Chunking
Practitioner-first engineering advice for engineers building retrieval augmented chatbots. Improve chunking, add production guardrails, and monitor failures.
Lakehouse for AI: Avoid a Second Vector Store, Be ADMP Audit Ready
Operations guidance for AI on a lakehouse: unify embeddings in one authoritative system and meet Malaysia ADMP rules.
Run LLMs on Private Data Safely With 3 Compliance Controls
Run LLMs on private data without leaks. Use retrieval-first or private hosting and apply three controls—sanitize inputs, enforce retrieval-time access,...
bottom of page