Naitik Chandak

I take AI agents from pilot to production, and make tracking something you can trust.

What do you need help with?

The demo worked, so ship it. Ship it when the test, the approver and the limits are in place.

Read the proposalBook a 30-minute conversation

Naitik Chandak. Founder of Beneath Analytics, based in Ireland.

Proposal

From pilot to production

The problem

Most agents stall after the pilot. The model is rarely the cause. Nobody agreed what good looks like, who signs off, or what the agent may touch.

What I would do

  • Agent readiness sprint

    2 to 4 weeks, fixed scope

    You leave with a ranked list of agent use cases, the evaluation plan and guardrails for the first one, and a roadmap your engineers can start on.

    Discuss a sprint

  • Fractional AI product lead

    Part-time, monthly

    I join your product team to own the agent roadmap, write the specs, set up evaluation, and work with engineering through launch and after.

    Discuss a retainer

Proof

  • blayerAI: an agent that writes analytics changes and opens a pull request. It never merges. A person approves every change.
  • InsureIQ: an insurance policy explainer that answers in plain English.
  • Led analytics delivery for a global ecommerce programme at Henry Schein, with engineers, analysts and consultants in one pod.

See the products I built

How it runs

  1. 1

    Scope

    Pick one job the agent does, and what it must never do.

  2. 2

    Guardrail

    Limit what it can read, say and change.

  3. 3

    Evaluate

    Write the test set and the pass mark before building, so good is agreed up front.

  4. 4

    Ship

    Release behind human approval. Relax it only as evidence builds.

  5. 5

    Monitor

    Watch for drift in production and alert a named owner.

Built

Built, not just advised

Products I designed and built. Each is shown as the decisions behind it. Status is stated plainly.

Early accessblayerai.beneathanalytics.io

An agent that implements, tests and validates analytics tracking.

Problem
Analytics implementations ship and are never validated. Tags break and events drift until the data is already wrong.
Key decision
The agent writes the implementation and opens a pull request. It never merges. A person approves every change.
Guardrails
Every change is checked for structure, required parameters and data types, then compared across platforms before release. Raw event data stays in the customer's environment.
  • Does:Writes the implementation and opens a pull request
  • Never:Merges anything without a person approving it

Track record

Track record

9+ years across software, analytics, product and founding. Five roles, from engineer to founder.

  1. 2026to now

    Founder and Managing Director, Beneath Analytics

    Relaunched around agentic AI product work. Building blayerAI and TalPool, and advising companies on getting agents into production.

  2. 2023to 2026

    Ecommerce Product Manager, Henry Schein

    Led analytics delivery for the global ecommerce transformation programme, with a cross-functional pod of engineers, analysts and consultants.

    • 12x more organic search impressions, from the technical SEO analytics work
    • Consent management rollout across EMEA, with server-side GTM and consent-aware tagging
    • Consolidated legacy reporting into GA4, BigQuery and Looker
    • One analytics intake and delivery process across 6 EMEA markets
  3. 2020to 2023

    Founder and Founding Product Manager, Beneath Analytics

    Product and retention strategy for an analytics platform used by US and EU clients, and analytics work for more than 50 enterprise clients.

    • Client retention up 15%, with a pod of 10+ across product, engineering and analytics
    • Active platform usage up 3x in two years
    • Reactivation campaign conversion up 22%, from lifecycle triggers
    • Return on ad spend up 18%, from real-time attribution dashboards
  4. 2019to 2020

    Customer Success Manager, Tatvic Analytics

    Managed analytics clients and a team in Ahmedabad, working in Google Tag Manager and Google Analytics.

  5. 2017to 2019

    Software Engineer, then Business Analyst

    Started as a software engineer at Datagres Technologies, then moved to business analysis at eProcurement Technologies and Simform.

About

Systems that look right but quietly are not.

Naitik Chandak

I trained as a software engineer, spent years making analytics trustworthy for enterprise teams, then moved into product. I started Beneath Analytics in 2020 and came back to it in 2026 to work on agents.

The through-line is the same: systems that look right but quietly are not. Agents are the newest and least checked of them.

I live in Ireland and have Stamp 4 permission, so no sponsorship is needed.

Contact

Start with a conversation

Thirty minutes to talk through what you are trying to ship. If it is not a fit, I will say so and point you elsewhere.

Book a 30-minute conversation

Or write to naitikchandak@proton.me, or find me on LinkedIn.

What happens next

  1. 1You get in touchBook a time, or send the form.
  2. 2I replyWithin two working days.
  3. 3We talkThirty minutes about what you are trying to ship.

I use your details only to reply. Privacy notice