I build the tools I write about.

General-purpose automation tools like n8n or Zapier are built for developers and generic business workflows. CA firms and finance professionals have no tool built around their actual work: GST portals, income tax filings, TDS compliance, bank statements, and transfer pricing.

Built Vega from the ground up as a node-based module builder where logic and UI are designed together, compiled into reusable modules, added to a dashboard, and shared across devices including mobile phones on the same network.

A working desktop application in beta with purpose-built nodes for GST Portal, Income Tax Portal, TRACES, Bank Statements, NSE Data, Google Maps, AI Agents, Web Automation, and more, with every node demonstrated through real workflows shown in this portfolio.

As Python automations multiply, managing dozens of scripts becomes chaotic, with no unified interface, no way to hand them to non-technical users, and no structure for tracking what does what.

Built a platform where an Excel file defines the entire UI: sheets become entities, sections become tiles, and rows define the input elements. Each tile connects to a Python script and runs it with the configured inputs through a clean browser interface.

Built on personal initiative during employment at a Big 4 firm, adopted and used widely across multiple teams for managing compliance and accounting automations without any technical knowledge required.

Selecting comparable companies for transfer pricing requires applying multiple financial filters across thousands of companies, then qualifying each one on business description, a process that could take days manually.

Built a web application with a local database of 50,000+ companies, configurable quantitative filters for standard transfer pricing parameters, and AutoFAR integration to automatically fetch FAR descriptions for companies not already in the database.

A working transfer pricing search tool built and used during professional practice, reducing a multi-day comparable selection process to a guided, filterable workflow.

Editing script-based and talking-head footage means manually scrubbing hours of takes, finding the good ones, and cutting filler and repeated attempts by hand. The mechanical work dominates the entire edit.

Built a browser-based editor that transcribes the footage, aligns it to the intended script with an LLM, and automatically assembles a cut that keeps the best take of each line and drops the repeats — with editing that stays transcript-aware: split, extend, and adjust any segment against the words.

In active development. The transcribe → script-align → auto-cut pipeline runs end to end; per-segment editing and export are being refined.

Tracking progress through large structured goals like thousands of exam questions, months of gym sessions, or long reading lists, has no good middle ground between a simple checklist and an overcomplicated app.

Built a PWA where any activity can be broken into chapters and steps, each marked complete with a tap, with importance flags, per-step notes, priority-based points, progress filters, and full backup and sharing via a username code.

Live at tapas.souravsensei.com — a free, installable tracker for study, gym, pills, and any task, shipped from the original Gamify prototype.