Project summary

I am a first-gen undergrad, coming from a humble background in Nepal, studying with a full scholarship at Rollins College in Florida. I have taken research initiatives myself to contribute to AI Safety; my first-author paper "Measuring the Integration of Adversarial Machine Learning into Practice Through Artifact-Anchored Methodology" (https://openreview.net/forum?id=6WbZR1ID6q) was selected for an oral presentation at ACML 2026 in Melbourne, Australia (Dec 1 to 4). We introduce AdoptionTracer, which measures how fast ML security research gets into production (~8.1-month median lag). Funding this trip will allow me to present these empirical AI safety metrics.

An in-person presentation is mandatory for the paper to appear in PMLR. My undergrad school has helped with costs for registration/visa; I am now left with flight costs (US to Australia) and staying in Melbourne (5 days) on my own, which I am unable to fund myself. I would really appreciate any support.

What are this project's goals? How will you achieve them?

Actions & Steps: I'll use the grant to immediately book required flights from the US to Australia, and back; deliver the oral presentation; help bridge the gap between theory and practice in AI security through our paper's method; network with researchers working on robust AI and safety evaluation; and potentially meet with an advisor for my future PhD or nonprofits working in the field.

Impact & Path to AI Safety: Our method, AdoptionTracer, provides the first empirical mechanism to track how security techniques translate into real-world practitioner tools. Presenting this live connects evaluation metrics directly with downstream builders.

How will you know it worked?

-Being present in person in Melbourne, Australia, at ACML (this is mandatory for our paper to appear in PMLR)

-Delivering the oral presentation on AdoptionTracer to the broader ML community

-Meeting with potential PhD advisors, organisations, or non-profits working in AI Safety

How will this funding be used?

Round-trip flight from Orlando, US to Melbourne, Australia about $1,500 (not booked yet). Additional costs of hostel and food in Melbourne for 5 days ($300).

Who is on your team? What's your track record on similar projects?

I am the main author of the paper and am responsible for presenting in person.

Website: https://madhav.club · LinkedIn: https://www.linkedin.com/in/madhav-khanal3145/

Before I was 20:

- Authored Abichal Karmayoddha, a bestselling Nepalese novel on homeless children

- National Team, 53rd International Physics Olympiad

- Full-ride Alfond Scholarship ($350k) for undergraduate study in the US

Now at 21:

Research Execution & Publications:

-In 2026, I wrote 5 first-author papers accepted or under review, including an oral presentation at ACML 2026 (PMLR) on adversarial ML adoption, an accepted paper at IEEE S&P (MetaCRISP) on LLM coding agents, ACNS 2026, and a submission to ICLR 2027.

AI Safety Involvement & Lab Experience:

- As a 2026 CMU RISS Scholar at the Auton Lab, I engineered Blindfold, a pre-deployment evaluation framework for DARPA's vision-language models and training-free late fusion.

- As an AI Safety Fellow with SPAR and SaferAI, I contributed to LLM forecasters for cyber risk models. I am also involved with

-I am also connected with broader AI Safety communities like Bluedot, AI Safety Nepal, etc.

What are the most likely causes and outcomes if this project fails?

If I cannot pay for the flight or fail to attend in person, we won't get to present the paper. Given my financial situation as a first-gen undergrad from humble backgrounds in Nepal, and with December approaching, flight costs will only go up, which doesn't help.

In the worst case, our work for the last few years won't be able to reach the broader AI Security community, and the paper won't appear in the Proceedings of Machine Learning Research, as presenting in person is mandatory.

How much money have you raised in the last 12 months, and from where?

BlueDot Impact: $1,950 Rapid Grant for my October 2026 trip to EDPM and IAPP (report sent). I applied to EA Funds for the remaining $1,800 (pending). School helped with mandatory non-student full registration ($700) and a visa for a Nepalese citizen ($300). BlueDot declined a second request for this trip. I have also emailed the program chairs requesting if there's any volunteering (not listed on website). If they respond, or if EA Funds or anyone else covers the flight, I will close this project or return the overlap.