Ola Krutrim GPU Pipeline: Automating an End-to-End Sales Workflow

Krishnam joined Ola Krutrim at a point where the GPU sales process was entirely manual and fragmented. Inbound demand was growing, but the team had no reliable way to track leads, score intent, or prioritize outreach. He was handed a broad mandate: automate and simplify the pipeline.

He started by mapping the current state end-to-end, sitting with the sales team to understand where leads were being lost and where the most time was being wasted. The first bottleneck he identified was intent classification: the team was treating every inbound email the same, regardless of how likely the sender was to convert.

His first build was an intent classifier that ingested production emails and assigned each one a confidence-interval score across purchase likelihood. Running the classifier on real emails surfaced a key problem: a large share of inbound requests were too vague to score reliably. Rather than discarding those leads, he designed an automated questionnaire that engaged the sender, captured structured requirement data, and fed it back into the scoring matrix.

The questionnaire served two purposes. It kept potential buyers engaged without requiring any manual effort from the sales team, and it generated the structured data needed to score previously unclassifiable leads. With the classifier and questionnaire in place, he built a pipeline dashboard that gave the team a live view of deal count, confidence distribution, and attributable revenue potential.

He owned the full stack: problem framing, classifier design, questionnaire logic, and dashboard delivery.