building ai for vcs, rethinking nexora

i’m splitting my time between my sde job, uni, and this thing i’m calling nexora—though i’m not sold on the name yet. work’s been intense, and it’s got me questioning what i want to build next. here’s what’s up. at my job, i’m working on a deep research engine for venture capitalists. it’s ai that digs through crazy amounts of data—reports, pitch decks, market trends—to surface insights vcs need to make bets. it’s not just googling; it’s like giving vcs x-ray vision into startups and markets. i’ve been grinding on ml models that have to parse messy, unstructured data, like pdfs or half-baked founder bios. last week, i was stuck on a python script that kept choking on weirdly formatted funding rounds—think csvs with missing fields or random text dumps. i spent a whole night rewriting how we clean the data, using some regex hacks and a new preprocessing pipeline. when it finally spit out clean insights, like a startup’s hidden traction, i was pumped lol. it’s wild to build ai that can think like a vc but faster, and it’s showing me how to craft systems that cut through noise. that’s messing with my head for nexora. i’ve been saying it’s about fixing the startup scene—linking founders, coders, investors who’d never cross paths. but this week, i read about this tech design thing called “jobs to be done,” where you focus on what people are trying to achieve, not just what they say they want. it’s got me rethinking. apparatus.ai helped small businesses, but we built some features just cuz we could, not cuz users needed them. for nexora, i’m wondering: what’s the real “job” for startups? maybe it’s not just connecting people but helping them find the exact resource they’re missing—a coder for a prototype, an investor who gets their niche. i’m not sure if nexora’s a platform with a slick ui or something leaner, like an api vcs and founders plug into. rn i’m prototyping on my linux box, using python to test ai matching and js for a quick frontend to see the results. i got a script to pair fake founders with investors based on pitch keywords, but it’s rough—crashed when i fed it bad data yesterday. still, it’s got me hyped, even if i might ditch the name. uni’s a grind, keeping me up late with assignments and group projects where i’m basically the only one coding. i’m hacking nexora in my room, surrounded by empty coffee cups and a million browser tabs—stack overflow, ml papers, startup blogs. apparatus was my first shot, and it landed, but it was small potatoes. this time, i want to build something that shakes things up, but the uncertainty’s real. is it nexora? a new name? a platform or a tool? i’m cool with not knowing yet. that jtbd framework’s stuck with me—it’s like, figure out the job first, then build. i’m trying to do that, one script at a time. no clue where this is headed, but i’m having fun messing with it. work’s teaching me ai can do more than i thought, and i want to bring that to startups. if you’re into ai or the startup game, hit me up with what you’re working on—i’m curious. i’ll post again in a couple days when i’ve got more to share. gotta bounce back. — ciao, basab