About Slake

Workflow systems that use AI within clear limits.

Slake is Anish Jhaveri's solo software practice for B2B teams that need search, routing, reporting, dashboards, or AIOS workflows layered onto the tools and databases they already run.

Query the facts Structured records through SQL or NoSQL queries, APIs, and permissions.

Retrieve the context Notes, PDFs, logs, comments, tickets, and source docs.

Render the output Constrained JSON, dashboards, review queues, logs, and handoff notes.

Review a workflow Try the RAG demo View work samples

From messy input to usable system.

Build model

Map the drag

Inputs, tools, rules, exceptions, owners, and the business reason to build.

Build the path

Search, classify, extract, route, sync, alert, review, and report where it helps.

Hand it over

Source code, setup notes, logs, dashboards, and enough context to maintain it.

10+ years Apps, APIs, data, deployment.

Live demo RAG with cited source chunks.

Build notes ScoutLocal product workflow.

What this is good for

When the work is too specific for SaaS tools.

AI search

Internal docs, product specs, support content, PDFs, or API references become searchable with citations and source-aware answers.

Workflow software

Leads, invoices, onboarding steps, support tickets, and CRM updates move through clear rules, review paths, and tool integrations.

Internal tools

Dashboards, admin surfaces, review queues, and reporting views give the team a real interface instead of another spreadsheet ritual.

MVP builds

When the problem is bigger than an internal workflow, I can scope and build the first usable product path with auth, data, core flows, and deployment.

Working examples

See how the work is built before you start a conversation.

Live RAG demo

Ask questions against Stripe implementation docs and inspect the cited chunks behind each answer.

Marketplace search build

Notes from a real product build involving natural-language search, auth, data modeling, and founder workflows.

Implementation judgment

Practical breakdowns on when to use an LLM, when deterministic logic is enough, and when custom code is not worth it.

How I work

Make the workflow easier to run.

Scope the workflow

We identify the input, decision points, edge cases, authoritative records, and success metric before talking about a build.

Keep humans in the loop

Good systems expose uncertainty. They show what happened, what was used, and where a human should review.

Build for ownership

The deliverable should be maintainable: source code, setup notes, logs, and enough context for your team or another developer to extend it.