the anti job boardactive 5mo ago
Vishnu Manoj
MLE Intern @ Human Archive (YC W26)
Summary
Vishnu Manoj is an AIML engineer with a strong background in building and deploying machine learning and computer vision systems. He has demonstrated expertise in developing production-ready AI solutions, optimizing inference processes, and integrating multimodal data streams, making significant contributions to various projects and internships.
Looking for
Skills
Preferences
Open to
San Francisco
The story
What Vishnu is looking for
Cutting through the clutter. I like my autonomy when moving through problem statements that are seemingly open-ended, my whole Motto is "Are we 1% closer to solving the problem today than we were yesterday?". Apart from that I enjoy the process of building a product from the ground up.
Why Vishnu is exploring
I’m a recent Master’s graduate in AI from Duke, currently based in NYC. I’ve always been someone who enjoys solving hard problems, which is why I pursued a mix of experiences across startups, research labs, and larger companies — including internships at JPMC and a deeptech startup. I’m currently working as an MLE at a YC W26 company. What’s pushing me to look right now is honestly the state of the entry-level market. Many roles ask for 2–3 years of experience even when they’re labeled entry level. I know I can build strong solutions and contribute quickly, but the traditional filters don’t always reflect that. So I’m looking for teams that value problem-solving ability, ownership, and learning speed; not just years on paper.
What won't fit on a resume
A lot of the most interesting things I’ve built don’t really fit neatly on a resume; they usually start as rabbit holes. For example, recently I built EgoCut, a tool to clean egocentric video data. Anyone working with head-mounted cameras knows that hours of raw footage are mostly unusable — blurry frames, idle time, repeated actions. I built a small pipeline that automatically filters useful segments and identifies repeated cycles so you can extract good training data much faster. No one asked me to build it — I just got curious about the problem. Demo: https://x.com/vishnutm244412/status/2025984677182403045?s=20 Repo: https://github.com/calicartels/EgoCut Another project that stuck with me was Blind.ai, which started as a 2-day hackathon project — a visual aid app for people with vision impairment that could read text, detect objects, recognize currency, and trigger SOS alerts using gestures and voice. We actually lost the hackathon, so I didn’t think much of it. But six months later I was volunteering as a scribe at an NGO helping visually impaired individuals read and write. They were short on volunteers, so I tried using the app to help them read documents. That eventually turned into a research project on OCR for an indigenous language and my first paper. It also taught me the difference between code that works in a demo and systems that have to work for real people. I also had a short entrepreneurial stint building Traceway, a GitHub-as-a-resume platform that tried to show developers through their code and project history instead of static resumes. It wasn’t a huge success, but it reached 1000+ concurrent users at its peak and taught me a lot about building products people actually use. More recently, I built a voice agent called Consilience for a Duke/OpenAI research project. The interesting part wasn’t just the models; it was encoding social behavior into the system so the AI behaves like a good facilitator: answering questions when asked, quietly monitoring discussions, and only speaking during natural pauses instead of interrupting people. Most of these projects started the same way; noticing a problem, getting curious, and trying to build something useful.
Work & Projects
Links
The Anti Job Board
Skip the application black hole. Get warm intros to founders hiring now.
Create your profile