loglens.labs
Loglens Labs Built by Gokul Anantha
Back to home

Gokul Anantha Founder, Loglens Labs Bengaluru

I build enterprise AI that runs inside the customer's own systems, so sensitive data never goes to an external model.

Background

That conviction didn't start with Loglens Labs. It came from years of CRM, front-office and enterprise middleware work, building the systems that had to sit next to sensitive data rather than send it elsewhere.

As Head of Product for SAP Intelligent Notification 365, I owned an embedded decisioning engine that chose channel, timing and failover for enterprise messaging across email, SMS, push and WhatsApp, deployed behind SAP Cloud for Customer and SAP Marketing Cloud. It processed more than a billion messages and helped grow annual recurring revenue from around $200M to $350M, and was named CTIA 2015 top 5.

Later, through Workato consulting, I built integrations that connected 29 enterprise apps, including Workday, Microsoft Dynamics 365 and BambooHR, to automated gifting workflows on Postal and Sendoso, so that employee, customer and partner events across those systems could trigger the right gift at the right moment, instead of staying invisible to the people running the program.

The pattern repeated every time: the data that mattered most was locked inside other systems, and the AI or logic that used it well had to be explainable and sit right next to it, not call out to it.

What I'm building

Loglens Labs is where I test that idea end to end, across different problems and company sizes.

LogloopLive with customers — adaptive nurture for B2B revenue email
GTM360, Collab, Quotation AI, cold-start dataIn progress — from internal testing to active build

See the full project list on the homepage →

What I believe

01

Narrow, high-volume operational tasks suit small, specialised models.

Scoring an intent signal, flagging a data conflict, deciding a quotation: these aren't general intelligence problems. A model trained for one job, at the volumes a business actually produces, beats a general model on cost and fit.

02

Data sovereignty is an architecture choice, not a compliance box.

Every call to an external model is a data transfer. When the data is sensitive and the inference can happen locally, the model should travel to the data, not the other way round.

03

AI should explain its outputs, so people trust and use it.

A score or a decision without its reasons is noise. The people who act on a model's output will only keep using what they can question, so every model we build returns its reasoning alongside its answer.

Get in touch

If you have a customer problem, a research question, or interest in a design partnership, I'd like to hear from you.