interview

Tell Me About Yourself?

• Graduating in a month as AI student at….

• Love building things from scratch - startups, coding projects

• Curious about how technology can transform industries

• Use AI to create meaningful change - not just impressive algorithms, but solutions that help people

• Want to make traditionally inaccessible things (like banking) more accessible through technology

• Hobbies….

"Walk me through your Brain Tumor Detection project - what challenges did you face and how did you solve them?

• Built neural network using TensorFlow/Keras

• Classified brain tumors from CT scans

• Goal = Achieve high accuracy for medical diagnosis support

• Challenge = Discovered serious data bias issues - model = unreliable across diff scan types

• I;

  • Analyzed dataset systematically to understand bias

  • Researched class balancing techniques + data preprocessing methods

  • Retrained model with balanced dataset

• Result = Achieved 92% accuracy + 15% faster training time

Why are you interested in OakNorth specifically, and how do you think AI can improve banking?

• Why OakNorth?

  • Mission of barrier-free banking resonates with my startup experience

  • Recent OpenAI partnership shows you're serious about AI innovation

• AI applications in banking:

  • Credit assessment: AI can analyze alternative data sources beyond traditional credit scores

  • Process automation: Streamline loan applications + reduce manual work

What are your weeknesses?

"Can be too hard on myself"

• overly self-critical

• set very high standards

• Double-edged sword: Good (self-reflect + constantly improve) Bad (Mentally exhausting + burnout - miss bigger picture)

• Eg: After 92% accuracy on Brain Tumor detection = very fixated on not getting higher like 96%

  To overcome this;

  • Peers perspective when too critical

  • Step back + see overall progress, not individual flaws

What's your biggest strength?

Problem-solving + Persistence

• Eg: Brain Tumor Detection Project - Discovered data bias issues.

• My approach:

  • Researched multiple solutions before choosing best

  • Kept going through the multiple failed models knowing that I will eventually meet the desired accuracy

• Crucial in AI/banking - can't cut corners

• Reliability + Accuracy = Crucial

1. "What are the latest AI trends you think could impact banking in the next 2 years?"

• Generative LLMs: Automating document processing, customer service + financial report generation

• Real-time risk assessment: AI models that continuously monitor + adjust risk in real-time

How do you stay current with AI developments and research?

• Academic sources: Follow key AI conferences like NeurIPS, ICML, and read papers on arXiv

• Experiment with new AI tools and frameworks in personal projects

• Networking with AI professionals + attend virtual conferences

What's your view on how LLMs like ChatGPT could be used in banking?

• Customer support using chatbots that understand complex financial queries

• Document analysis: Automatically extract insights from loan applications, contracts, and financial statements

• Compliance: Help interpret + ensure adherence to complex banking regulations

How would you use Large Language Models to improve customer service in banking?

• 24/7 intelligent support: Handle complex queries beyond simple FAQs

• Multilingual support: Serve diverse customer base in their preferred language

What are the risks and benefits of using LLMs for financial advice?

Benefits:

• Accessibility: financial advice for customers who can't afford human advisors

• Personalization: Tailor advice to individual financial situations

Risks:

• Regulatory liability: Who's responsible if AI gives bad financial advice?

• Hallucinations: LLMs might generate plausible but incorrect financial information

What AI techniques could help detect fraudulent transactions?"

• Anomaly detection: Identify unusual spending patterns

• Machine learning ensembles: Combine multiple models for better accuracy

• Feature engineering: Use transaction timing, location, + amount patterns