Balbir Singh Start a conversation
Menu

Case study

Raptor Labs: Hercules + Industrial AI

A program portfolio that combined AI-powered catalog onboarding with a conversational buying interface for industrial products.

Context

The problem was never just search. It was the entire product-data system behind industrial discovery.

Raptor Labs sat at the intersection of product data, ecommerce search, and AI execution. The wedge was straightforward: make impossible-to-onboard products commercially searchable and purchasable.

Hercules handled product mapping, cleaning, and standardization. Industrial AI extended that foundation into an LLM-powered discovery experience.

Revenue signal

£4.5M+ generated from products unlocked by Hercules

Scale signal

15x product growth and 8x spec growth

Efficiency signal

99% reduction in grouping effort

Problems

What had to be solved

Industrial ecommerce breaks when source data is inconsistent, discovery is shallow, and product teams treat every symptom as a separate fix.

Hercules

Catalog onboarding needed to become an AI and data-engineering system.

Supplier documents, fragmented mappings, and inconsistent product structures were limiting both assortment growth and revenue capture.

Industrial AI

Search behavior needed to evolve from filters and clicks to guided conversation.

Industrial buying often starts with partial knowledge, alternate part needs, and contextual follow-up questions. The interface had to reflect that reality.

Approach

How the systems were shaped

The work combined product framing, architectural ownership, and coordination discipline.

01

Frame the commercial wedge

Translate vague AI enthusiasm into a concrete operating problem with clear upside, urgency, and owners.

02

Build the data spine

Design the pipelines, product signals, and intelligence layers that make the system trustworthy under scale.

03

Ship through coordination

Align stakeholders, sequence the roadmap, and keep execution close to measurable business movement.

Execution

Technical depth that mattered

The resume signals are specific because the systems were specific.

Hercules stack

  • PDF-to-structured-data pipelines with layout extraction.
  • Fine-tuned DIT and relation extraction using spatial features.
  • Product mapping, cleaning, standardization, and grouping automation.

Industrial AI stack

  • Query decomposition, rewriting, and intent classification.
  • Named entity recognition and context management.
  • Response generation and alternate-parts discovery systems.

Contact

Work with Balbir

Useful for teams that need one leader who can connect product strategy, AI architecture, and operating discipline.

Contact

Start a conversation

Best for leadership roles, advisory opportunities, or product and data problems that need both strategic clarity and execution depth.

hello@balbir.netbalbir4email@gmail.com
Typical topics

Leadership roles · advisory work · search, data, AI, and execution systems

Best starting point

A short note with the problem space, team context, and what needs untangling.

Best fit

Leadership roles, advisory engagements, and AI programs where business leverage depends on system design.

Especially relevant when the real bottleneck is not a feature request but the structure underneath it.