If you're not sure it can be built, or whether to build or buy, I can help you find out, then lead the build.
I've been building software for 27 years, from GPU speedups to computer vision and audio, long before any of it was called AI. So I can usually tell quickly whether something can be built, and what it will really cost.
90-minute call, $500. Refundable in one click.
- A demo that worked on sample data
- Vendor quotes and a build estimate
- A quote for GPU servers
- Measured on your own data
- Build and buy, side by side
- Running costs, at your volume
What I usually find
Usually, it's new to the team
The team often has the right intent, but not yet the experience for the decision in front of it. Nobody in the room has built this exact thing before, and the best person for a hard problem is someone who has already solved it.
First, a small prototype
If the answer isn't clear on the call, we guide your team to build a small prototype on your own data, and help you read what it shows, before the budget moves. Real data is messier than the demo's.
Bought the CRM, built the portal
At a revived heritage travel brand, we moved a 40-person sales team onto a new CRM, bought off the shelf. And an in-house portal replaced several paid travel-booking tools in about three months, built around how their team actually works. It cut both the licensing and the running costs.
A note from me
The standard test images for compression research were too small, too old, and only 8-bit precision. In 2007, I put together a new high-resolution, high-precision set, each image picked to stress the algorithms in a different way. I test your prototype on the most appropriate data.
- Sachin
Some of the work
Compressing an 8 megapixel image with a proprietary algorithm took 3 to 4 minutes on iOS and Android. At Rawzor, we moved the same algorithm onto the GPU, and it took around 10 seconds, 94.7% less time. We used OpenGL shaders instead of CUDA or OpenCL, so it stayed portable across desktop and mobile.
In 2013, I developed and trained computer vision machine learning models for detecting and tracking fish species in open waters. Keeping detection and tracking separate made it far more efficient: up to 100 objects at 30 FPS on a 3.1 GHz 4-core Intel i7, using OpenCV.
A New York based VoIP company had spent nine months on an AI project that was moving at a crawl: self-hosting ASR models and LLMs for transcribing and summarizing calls, to keep costs down at scale. I recruited the right specialists, and it shipped in two months. I also drove the GPU batch-processing optimizations.
Clients stay unnamed here. I'm happy to walk you through any of these on a call.
How I work
What I look at first
What you're trying to achieve, and where things really stand. The quality of the work so far, and how the ground reality compares with how it feels to you. That can go either way: sometimes the work is worse than it looks, and sometimes it's much better than it feels, simply because it isn't finished yet. The skills the job needs against the skills on the team. Whether your timeline and budget estimates are realistic. And the gap between what I hear from leadership and what I hear from the people doing the work. That gap usually tells me the most.
Progress everyone can see
Status updates that say what's really done, in enough detail to act on, and a working build you can try every week. A lot of the misunderstandings I see start with updates that skip the details.
People who've solved it before
I look for people and tools that have already solved the exact problem you're struggling with, not just a good developer, a good marketing team, or a good CRM. A job post that says “great machine learning engineer” vs “a machine learning engineer who has worked on audio models, and plays a musical instrument” can turn a 3-month project into a 3-week sprint.
A team that doesn't need me
Knowledge moves to your team, and if you need a full-time CTO, I help you find the right one. By then, no part of your product should depend on one person, including me.
How we'd start
Everything starts with one paid call. By the end of it, we'll both know whether there's more to do, and which of the other two options fits.
A first call
- 90 minutes with me, working on your actual problem.
- Real answers and clear next steps on the calls themselves.
- Up to two follow-ups, with you or the people on your team.
- If it wasn't worth it, one click refunds the full $500.
The right first step for nearly everyone.
Some teams take the solutions and run with them on their own. That's a perfectly good outcome.
A fixed-price review
A close look at the question and at what you have so far, over two to three weeks.
- A small prototype, built and measured on your own data
- The cost of each option: build or buy, your own servers or an API
- Solutions to the problems we find, worked out together with you and your team
Your fractional CTO
A few hours a week on the hard technical calls, and on getting the right people to build it. The monthly range depends on how much of my time you need.
- Architecture, hiring, and vendor calls
- Finding specialists who have already solved your exact problem
- A planned hand-off to your own team
Bootstrapped and pre-revenue? Book a 40-minute Founder Call for $200. It doesn't include follow-up calls, and the same one-click refund applies.
Fair questions
Because it's working time on your problem, and you get real answers on the call itself. It also covers up to two follow-ups, including calls with people on your team if needed.
Sometimes, for a tricky performance problem or a code review. Mostly I make sure the right people are writing it, and that what they build is what you actually need.
Then I'll tell you plainly, and show you why, before the money moves.
A sample of your real data, whatever you've built or tried so far, and time with the people who know the problem best.
No. I take no reseller fees or commissions from any vendor, so I'm truly on your side, with only your best interest in mind.
Book your first call
Pick a time that works for you. Bring whoever on your team knows the technical side best.
$500 for the first call
- 90 minutes with me, working on your actual problem.
- Real answers and clear next steps on the calls themselves.
- Up to two follow-ups, with you or the people on your team.
- If it wasn't worth it, one click refunds the full $500.
Need more than one person? Rawzor is my firm, a small senior team of experts that works alongside your people. Explore rawzor.com →