About

I’m Subhajit. I help companies turn messy documents, business data, and internal paperwork into structured, usable information.

Most of my work is around document automation, OCR, and data extraction. I build systems that process scanned PDFs, invoices, lab reports, contracts, and email attachments — the kind of documents where standard tools often fail because the layouts keep changing.

I usually start by understanding your business process first: what documents come in, what data matters, where errors happen, who reviews the result, and where the final output needs to go. After that, I design a solution that’s realistic for your budget, data quality, and timeline.

How I work

#

I start with the simplest approach and iterate from there. Over-engineering is the most common failure mode in document processing — reaching for LLMs when regex would do, or building custom infrastructure when a single API call covers it.

My pipeline development follows the same sequence every time:

Define the schema first. If I can’t describe what clean output looks like, extraction will always be unreliable.

Establish a baseline with rules. A deterministic check tells me exactly where the hard cases are before introducing any probabilistic tool.

Introduce LLMs incrementally. If a simpler rule-based approach gets you 90% of the way, I’ll say so — and we skip the LLM.

Apply layout detection last. Only where variation genuinely requires it.

If something can be done with a simpler approach, I will tell you. If an AI model is useful, I’ll explain where it helps, where it may fail, and how we validate the results before using them in a real business process.

What I do

#

Document Extraction Pipeline. I build systems that extract structured data from PDFs, invoices, lab reports, contracts, and email attachments. They handle layout variation and format drift — the things that break simpler solutions once you’re in production.

System Integration & Ongoing Support. I connect extraction pipelines to downstream systems: Excel, databases, APIs, internal tools. Retainer support is available as your document types change.

IDP Diagnostic & Action Plan. If your document pipeline is failing, I audit your current approach, identify the failure points, and deliver a written action plan with execution steps. Fixed price.

Results

#

• Cut manual data entry by 75% for a water consultancy processing lab reports across 10+ changing layout variations
• Maintained and expanded over a two-year retainer — the system became central to the client’s daily operations
• Extended the pipeline to process email attachments directly, removing a manual download-and-upload step

What clients say

#

“Thanks to Subhajit’s work, we are saving countless hours having to manually enter results into our own template.”

“Subhajit’s professionalism, meticulous attention to detail, and expert handling of our project were apparent from the outset.”

“He always delivers great results. His communication is clear, he takes feedback well, and he consistently finds effective solutions.”

Let’s talk

#

Book a free 30-minute call. Bring your documents. I’ll tell you whether automation makes sense for your case and what a realistic solution looks like.