Sid Venkatesh

hey, I'm

Sid

Venkatesh

My hair doesn't look this good.

Had a little help from Gemini.

Shortcut to projectsfor people who get straight to business.

Visit

Mechatronics engineer → ChatGPT enthusiast → AI powered IDE user → agent specialist.

None of this taught me anything about how to make a good product.

A good product isn't something cool and perfect. Learnt that the hard way.

Letting go of perfection is the hardest thing I have ever had to do.

— The guy who made this website

Mistakes that taught me

Ramwall GPT

Where do I start? This was meant to be a clone of ChatGPT, with access to the client's OneDrive, Outlook, and email, The agent could send emails, create documents, set schedules and tasks, with the full context of the business. The client absolutely loved the idea. It's awesome, of course. At the time it was just me working on the project. So I essentially single handedly wanted to build what Claude Cowork and Claude Code do, on a two month deadline. Ha! Ambition is not something you'll find me lacking in. I had actually got a working design going. The agent could complete complex tasks like "get me the latest meeting minutes, modify it, and send to Steve," with granular approval so people could see and sign off on each step before anything actually sent! Unfortunately, with limited resources and a tight deadline, I couldn't guarantee production-grade reliability. I couldn't finish the project and I found that the client still preferred using ChatGPT over this, as ChatGPT was shipping features faster than I could. It was slow, clunky, and showed my heavy bias towards Google infrastructure. The client ended up settling for smaller agentic workflows for specific tasks at their company.

StackFlutter web + Firebase hosting + Firestore DB

Ora

Now, this was an interesting one. I really got into AI music in this period. I was absolutely fascinated by the fact that AI could make music. The big players in the space were Suno, Udio, and Riffusion, of course. I wanted to own my own music generation platform, so this turned into about a two week stint of 4+ overnighters, UI and UX, fine tuning the model, and backend infra, all stacked on top of each other. I tried the best open source AI music models I could find, especially ACE Step at the time. I hosted the model on rented GPUs on RunPod, and had a working system and library of music going. I shared it with a couple of my friends, but no one could deny that the quality was nowhere near the big players. Even if mine was cheaper, it wasn't great. I had to retire it after a few months. Solid effort, though.

StackHTML, CSS, and JS + Cloudflare R2 for media storage + Firebase for auth and data storage

Alias

Now, I wouldn't go so far as to say this was a mistake. It was born from a solid idea, connecting customers to the right businesses and businesses to the right customers, as fast as possible. Where we went wrong was probably scope, strategy and execution. We tried to do too much, and tried to come up with a solution that fits every business and every customer. If you want an agent to buy a product or book a service, first you need to settle whether people even want to buy products from an agent. They might for certain necessary products. Then how do you compete with the titans like Google, Amazon, Shopify? With services, salons are different from restaurants, and both are different from trades. Can you provide the same solution for every business? We developed an MVP for this. Businesses got set up with a business agent that could help them enter their business info, add services, and manage the business through agent interactions, and vibe code their site. But that wouldn't be good enough to get businesses on the platform on its own. Customers aren't booking through agent interactions as much yet, and businesses need immediate value, not just a promise for the future. But... we have pivoted really well. Feedback from initial clients has been honest about what's missing, and that's pushed us toward a product that solves real problems first while gathering the data we need for agentic booking as we go. This will be exciting. Watch this space.

StackNext.js + Vercel + Python agents on Cloud Run

Not necessarily cool, but solved real problems.

Toastmasters Daily

This was not something I was expecting to become anything. My Toastmasters club at the time struggled with storing guest info and having it in one place, and we needed to trim how much paper we used for meetings. So I thought a site might be worth building, to solve the problem for my own club. Members from other clubs visited, really liked what they saw. In their words, they were like, "I like that, I want that, how do I get that?" So it actually became a product. Clubs and members gave feedback on what features they needed. I focused on shipping, and there were new features landing about every two days. One of those was a chat agent on the home screen called Toasty. It could answer the usual newcomer questions about how Toastmasters works, point people toward the nearest club from wherever they were, and help them join meetings as well. Now it's used by clubs in NZ. The aim is to keep expanding until clubs everywhere can use it. Quite a way to go.

StackHTML, CSS, JS + Firebase Hosting + Python agent on Cloud Run

Indosole

Indosole in Bali uses Shopify with Cin7. They had a lot of issues with Cin7. Stock take tools that worked with their account were missing, and reordering wasn't transparent or working to their needs. So we developed a site covering AI insights and reporting, stock take management, sales forecasting, and reordering, because their Shopify Cin7 flow still wasn't handling those parts. All of that together saves the team weeks of time they were handling manually before.

StackNext.js + Vercel + Firebase Auth + Neon DB

If you have a project in mind or just want to have a chat,

Let's talk.

Email me