Let's talk.

Whether it's a project, a role, or you just want to say hi, my inbox is open.

kadhar.mfn@gmail.com
AboutExperienceProjectsWritingResearchFree guidesToolsGames

© 2026 Kadharmoideen Fadurudeen. Made with care.

Kadhar
AboutExperienceProjectsWritingResearchFree guides
Tools
Say hello

Research & community

Questions I've studied, and ways I give back

I also write research papers about making AI safer and more useful for everyone, from people checking the weather to children who can't see a screen. Each one is a free preprint on arXiv.

Preprints

3 papers so far

All on arXiv
AI safety & evaluationNovember 3, 2026·arXiv:2611.05290

False Reassurance in LLM Severe-Weather Guidance: A Replayable Evaluation Against Real NWS Alerts

If an AI assistant can't see official weather warnings, will it tell people everything is calm while a storm warning is active? I replayed 117 real cases against five setups of a Claude-based assistant to find out.

What I found

Without access to real alerts, the assistant gave false reassurance in more than a third of warning cases. With real alerts and a neutral prompt, that dropped to zero, but prompt wording still mattered a lot.

Dataset, labels and code released openly

Read the paper: False Reassurance in LLM Severe-Weather Guidance: A Replayable Evaluation Against Real NWS Alerts
AI agentsAugust 5, 2026·arXiv:2608.07593

Weather- and Location-Aware Agentic Dining Recommendation: Leveraging LLM World Knowledge for Region-Sensitive Contextual Reasoning

People choose food differently depending on the weather and where they live: a rainy day means something different in Chennai than in Seattle. This paper describes an AI agent that checks location and weather, then reasons about local culture to suggest where to eat.

What I found

A working prototype showed an AI model's world knowledge can replace hand-written, region-by-region rules for this kind of recommendation.

Prototype built and deployed end to end

Read the paper: Weather- and Location-Aware Agentic Dining Recommendation: Leveraging LLM World Knowledge for Region-Sensitive Contextual Reasoning
AccessibilityJuly 24, 2026·arXiv:2607.22377

Kutti AI: A Voice-First, Offline-Capable Learning Companion with Real-Time Struggle Detection for Visually-Impaired Children

Most learning apps for kids are built around screens, which leaves out children who are blind or have low vision. Kutti AI is a learning companion that works entirely by voice, even without internet, in English and Tamil.

What I found

It notices when a child is struggling, from pauses and hesitation, and steps in to help. It runs on ordinary, affordable phones.

Built at the Half Baked hackathon

Read the paper: Kutti AI: A Voice-First, Offline-Capable Learning Companion with Real-Time Struggle Detection for Visually-Impaired Children

Preprints are shared openly before formal peer review.

Community

Helping other builders

Some of the most rewarding work happens outside a job description.

Hackathon judge

I judge hackathons, reviewing what teams build in a weekend and giving feedback on the idea, the engineering and the real-world impact.

IEEE member

Member of IEEE, the world's largest professional organisation for engineering and technology.

BCS member

Member of BCS, The Chartered Institute for IT, the UK's professional body for people working in technology.