Friends+Family DirectAI Monthly Update (August 2023)
Hey all!
Keeping you up to date with what’s new at DirectAI. To reiterate, we’re enabling anyone to build computer vision models that classify, detect, and track objects in images & videos, without training data. If you missed last month’s update, check it out here!
Last Month’s Rundown
August was primarily focused on recruiting a fall intern. We received 100+ applications for the position. We were heads down on building an effective recruiting process and a customized “super day” interview. We’re hoping to make an offer in the next few days.
Ultimately, we failed to meet our goal of meeting with 15 leads, largely as a result of lost time due to recruiting efforts. In terms of customer traction:
We published a blog post with and gave a talk to engineers at Ant Media. They’re gearing up for computer vision demos to customers at a conference next weekend.
We are exploring a deal to automate parts of the content moderation pipeline for another video streaming engine.
We are exploring a deal to perform semantic segmentation for a company that analyzes plants. This comes on the tail of several leads over the past few months explicitly asking (and expressing willingness to pay) for segmentation.
We added a number of fixes to our website demo and API that reduced unexplained errors & improved the user experience. In response to a couple of outages in late July, we have a playbook for quickly switching to a backup server. Uptime for our first full calendar month in operation was 99.8% and we expect higher going forward!
What We Learned
We ran a Google Ads experiment for a couple weeks to see what the cost of user signups & semi-qualified calls with leads would be.
Learning: Not an effective use of resources right now but it was useful to explore a potential bottoms-up sales tactic.
We realized that we often set goals at the beginning of the month that require certain assumptions to pan out (e.g. moving further in the sales process with a specific customer). We define specific tasks and quickly abandon them as the landscape changes. Plans are made to be thrown out, yes, but it’s pretty annoying to fail to accomplish something that’s written on the whiteboard.
Learning: We’ve decided to make our monthly goal very simple and specific - increase monthly recurring revenue. Given revenue is a lagging indicator, we can define a couple sub-metrics to game week by week (e.g. recurring API users, demos for prospective customers, public releases). And we can brainstorm/prioritize small tasks on the fly (every monday or every morning) to make those numbers go up. See YC’s podcast on this topic.
We overloaded on hiring this month. Our initial screening process wasn’t rigorous enough so I ran 15 minute interviews with too many people AND I made the mistake of using a calendly link that was open 9-5 M-F. Isaac, similarly, was booked solid running technical interviews.
Learning: Don’t spend more than 50% of a given week on recruiting tasks. We each lost ~2 weeks of productivity on non-hiring tasks due to overscheduling. We were significantly slower on responding to customer feedback and requests from prospective customers.
What’s Up Next
We have a backlog of potential customers that are waiting on us to demonstrate DirectAI’s inference quality on their data. We’re hoping to build cool stuff for as many of them as possible. We’re actively prototyping a semantic segmentation product (i.e. segment anything without the need to click) and the ability to define model classes with a couple images instead of solely with text.
Our AWS credits are expiring soon so we’ll be moving our GPUs over to Azure. The software we’re using to do this (Runhouse) will also address some of the root causes of previous production outages.
Onboard our first intern?!
Asks
Do you know anyone at the following companies? Please help me get in touch with them!
I have a forwardable blurb here and our shared calendly is here!
Kindly,
Ben
P.S. Check out this demo I made for hackers at PennApps! We’re giving out free credits for this weekend’s hackathon.