AI systems that do the work, and leave a record you can audit.

All eight service lines, in detail

• SourceTrail

  A desktop research agent that plans its own questions, searches the live web, verifies every claim against its sources, and writes cited reports.

  Python 3.12 · FastAPI · PostgreSQL 16 · Qdrant · Electron

• Francurial

  Browser-automation-as-a-service. Post one HTTP request and an isolated Chromium session carries out the task end to end, with a live link so a human can watch and take over.

  FastAPI · Node.js · Playwright · Redis Streams · PostgreSQL

  11 stars · 8 forks · MIT

• Sherlock AI

  A desktop app that analyses video-call interviews live — identity, gaze, transcription — and produces an evidence-backed report afterwards.

  Electron · React · SQLite · FastAPI · InsightFace

  MIT

• Gandiva AI

  A real-time desktop assistant that listens, reads the screen, and answers from your own documents — running locally by default.

  Electron · React · TypeScript · Node.js · WebSockets

  10 stars · 5 forks · MIT

Read how each one works

1. 01

   Discovery call

   Twenty to thirty minutes. What you are trying to do, what already exists, and what done looks like. No charge, and no obligation at the end of it.

2. 02

   Scope and proposal

   A written scope with deliverables, milestones, timeline and a fixed price. You decide with the whole picture in front of you rather than a number in an email.

3. 03

   Build in visible increments

   Working software at each milestone, not a status report. You see it running before it is finished, which is when changing direction is still cheap.

4. 04

   Handover

   Source code in your repository, deployed and documented, with a walkthrough for your team. Optional support after, as a retainer.

What we need from you at each step

github.com/h-a-r-s-h-s-r-a-h

Replies within one business day

What does a typical project cost?

It depends on scope, and anyone who quotes you before understanding the scope is guessing. We work three ways: a paid scoping sprint (small, fixed, and the low-risk way to start), a fixed-scope project, or a monthly retainer for ongoing work. A scoping call produces a firm written quote before you commit to anything.

How long does a project take?

A focused build — one agent, one automation, one retrieval pipeline — is usually 3 to 8 weeks. A full system with a product around it is 8 to 16. A model feasibility spike is 1 to 2. Each service page carries its own band, and the proposal carries a date.

Do we own the code?

Yes. You own the source outright, and it lives in your repository from the first commit, not ours. That includes the infrastructure definitions, the deployment scripts and the documentation.

Can you work with our existing team and codebase?

Yes, and it is often the faster path. We can work inside your repo against your conventions and review process, or build a service alongside it with a clean interface. We will tell you which one we think fits after seeing the code.

Do you only build AI?

No. We also build the product around it — API, database, auth, web app, mobile client — and on-chain systems where the ledger is the point. AI is what we lead with because it is the hardest part and where our public code is, but a model with no product around it is not something anyone can use.

Can you build the whole product, or only the AI part?

The whole thing. SourceTrail and Francurial are complete systems: queues, storage, auth, deployment, a desktop or web client, and tests. That is the same work, whether or not there is a model in the middle of it.

Have you shipped mobile or blockchain work we can look at?

Not publicly, yet. Our public code is AI, backend and cross-platform desktop — that is what you can read and verify today. We would rather say that than point you at something adjacent and let you assume. If those lines matter to your project, the paid scoping sprint is the low-risk way to test the fit before you commit to a build.

Do we need to train a model, or will an API do?

Usually an API does, and we will say so when it is true — it is cheaper for you and faster to ship. Training earns its place when a general model is too slow, too expensive at your volume, or not accurate enough on your domain. We establish that with a measured baseline before anyone writes a training loop.

Will our smart contracts be audited?

For anything holding real value, yes: we write and test them, then arrange an independent third-party audit before mainnet. It appears as a line in the proposal rather than as a surprise. We do not self-certify our own contracts as secure.

What happens after launch — do you support it?

Handover includes a walkthrough for your team, so you are not dependent on us. If you want us to stay on for monitoring, changes and the next phase, that is a monthly retainer. Both are fine; neither is assumed.

Can you sign an NDA?

Yes. Send yours, or ask and we will send one.

Which AI models do you use?

Whichever fits the project on cost, latency and privacy — we are not tied to a provider. Where data cannot leave your infrastructure, we use local models; Gandiva AI runs that way by default. The choice is made per project and written into the proposal.

What if we’re not sure what we need yet?

That is the normal starting point, and it is what the scoping sprint is for: a short paid engagement that ends with a written scope, an architecture, a timeline and a fixed price. You can take that document to anyone. If we are not the right people to build it, you will have lost a small fixed amount rather than a quarter.

How do we start?

Send a few lines about what you are trying to build. We reply within one business day, usually with a couple of questions, then a 20 to 30 minute call.