ken sugsavanvit

as a kid in bangkok i went to art school on weekends and drew for ten hours at a stretch. making a piece other people could stand in front of was one of the best feelings i knew, and it still is. at fourteen i picked up my dad's camera; i didn't know what i was doing, but i knew what i liked when i saw it, and the color-grading obsession never left (gawx and sleepy charlie are to blame).

somewhere in there i also started competing in speedcubing, which is still the cleanest hobby i have; you either got faster or you didn't, and the timer doesn't care how you feel about it.

covid lockdown. a senior took me on a jog and i fell into running; too far, honestly, in a way that stopped being healthy. climbing back to balance is where my one tattoo comes from, a line through a circle: the middle ground. reality usually lands somewhere between your brightest expectations and your darkest preparations, and finding the correct point takes real judgment, the way a doctor finds the right dose instead of splitting the difference.

i started lifting on december 14, 2022 and haven't missed a week since. half marathon in 1:31; bench 265 lb at 150 bodyweight. both, not either.

turned the camera into kenfilms: $10k+ in freelance revenue, a youtube channel past 2m views and 6,000 subscribers, cinematic promos for startups (cluely, captions AI, superprofile). along the way i taught video editing to 40+ kids of incarcerated parents, and a short film i made got showcased at a small festival; my family came to watch.

spent the summer of 2024 at CMKL (carnegie mellon × KMITL) as a research assistant, co-authoring two deep learning papers: parameter cycling, on training efficiency, and contextual encoding, on interpretability.

came to wesleyan as a freeman asian scholar: a full ride given to eleven students across asia.

started building hybridaf in may, solo, because the existing fitness apps make you pick between endurance and strength and i don't accept that premise. a next.js web app and a swift ios app, about 86k lines in eleven weeks; i leaned on AI coding tools for the typing and kept the architecture, the data model and every code review for myself. it's been on the app store since august.

in april i joined wesleyan's privacy tech lab under prof. sebastian zimmeck, where we're trying to work out what global privacy control should look like for AI agents. GPC already works on the web; the open question is what happens once an agent is acting on your behalf and handing your request off to other agents, because your opt-out has to survive every hop or it may as well not exist. we're building it over the model context protocol, and the prototype is public on github. in october the lab put out the paper, and i'm on it, second of seven authors: privacy preference controls for agentic systems, on arxiv and submitted to ACM CS&Law.

ran growth at spike from march to july: i scraped creator metrics off short-form platforms, wrote a UGC playbook from what the data said, and the videos made with it passed 500k views. then a summer back in bangkok as a data science intern at KBTG (kasikorn bank), modeling non-performing assets and real estate values to automate property appraisal; turnaround dropped by more than 99% and cost by about 85%.

back at wesleyan, and teaching now too: i'm a TA for COMP 211 under prof. victoria manfredi, running twice-weekly help sessions and labs for about 35 students and grading their C by compiling it and testing it against the rubric. i also tutor six hours a week at the SciT center, mostly helping beginners figure out why their code won't run. still at the privacy lab.

kenfilms keeps going on the side: paid reels for invideo AI, and @kenvizual, an instagram page about video editing that hit 1m impressions in its best month.

what i want long-term, i'm not sure yet; i just know i want to see people benefit from what i build.