Episode
S12 Bonus: Dr. Aqib Rashid, Glasswall
- Published
- Apr 9, 2026
- Duration seconds
- 1790
- Processing state
processed- Canonical source
- https://codestory.co/podcast/bonus-dr-aqib-rashid-glasswall/
Actions
POST https://stenobird.com/v1/public/podcasts/code-story/episodes/s12-bonus-dr-aqib-rashid-glasswall/transcription-requests
Idempotently request low-priority transcript generation for this episode.GET https://stenobird.com/podcast/code-story/s12-bonus-dr-aqib-rashid-glasswall.md
Read the agent-friendly Markdown representation of this episode resource.
Summary
Dr. Aqib Rashid explains how to transition machine learning research from a PhD thesis into a commercially viable cybersecurity product. He details the engineering rigor required to build scalable, trustworthy AI pipelines for malware detection.
Topics
- Cybersecurity
- Machine Learning
- Malware Detection
- Product Development
- Artificial Intelligence
- Software Engineering
- Startups
- MLOps
Highlights
- Main idea: Moving from academic proof-of-concept to a production-grade ML development lifecycle (MLDLC) requires rigorous engineering and automated experimentation
- Practical takeaway: To build a sustainable AI product, start with a repeatable methodology for one file type (like PDF) before attempting to scale across all formats
- Failure mode: Building impressive models that perform well on benchmarks but fail under real-world adversarial conditions
- Practical takeaway: Prioritize building products that earn user trust through explainability and usability rather than just technical complexity
- Main idea: Successful engineering teams should prioritize genuine curiosity about the problem domain over mere proficiency in specific ML frameworks
Chapters
1:00Defining the MVP: Dr. Rashid discusses the initial goal of using CDR telemetry to improve malware detection performance.6:40The Power of CDR: An exploration of Content Disarm and Reconstruction (CDR) as a foundation for file analysis.9:20Bridging Research and Production: The challenge of implementing machine learning models into a commercial product roadmap.12:00Achieving Commercial Viability: Balancing true positive rates with the need to minimize false alarms for enterprise customers.17:40Scaling the Training Pipeline: Building infrastructure capable of processing millions of files for model training in hours.20:30Hiring for Curiosity: Why domain knowledge and curiosity are more vital than technical skills when building a specialized team.34:20Advice for Entrepreneurs: A final lesson on building products that deserve trust and avoiding easily replicable AI features.