Skip to content
AKSHAY KANKAL

I build AI that enterprises can trust. And hardware that cuts power bills.

I lead engineering teams. We build AI assistants that big companies can trust with private information, so staff get a straight answer and never see a document they are not allowed to read. I also build hardware that cuts factory electricity bills.

Ready to relocate

Try the demo belowEmail me

Try it

Most AI demos show you an answer. This one shows you a refusal.

This is a working example of the kind of assistant I build. Choose a department, then choose a question. Now change the department and ask the same question again, and watch it refuse.

Step 1. Choose the department you work in

Step 2. Choose a question to ask

Made up documents, for illustration. No real company data.

What happens behind the scenes

  • FY26 travel and expense policyfinance
  • Supplier payment terms, EUfinance

Submit the claim in the expense portal within 30 days of travel, attach receipts over 25 euro, and your cost centre owner approves it.

How I think

What I believe about the work.

  1. The idea and the design should come from the engineer. AI should refine and speed up the execution. Engineers keep ownership of every decision and every line that ships.

  • 20,000

    Employees reachedBrose AI assistant, approximate

  • 50,000

    Messages per dayQuestions and answers, approximate

  • 76

    LocationsIn 25 countries

  • 50

    DevelopersUsing the AI guardrails, approximate

Built

Three systems I designed and built.

Each one had to work for ordinary staff who never asked for AI, and respect rules about who is allowed to see what, written long before anyone thought about AI.

01Brose

A company assistant 20,000 people actually use

An internal AI assistant for Brose staff across 76 locations in 25 countries. Ask it a work question, get an answer from the company own documents.

  • Answers from company documents for Finance, HR, Controlling and Shopfloor, rather than guessing from the open internet.
  • It does things, not just answers. Staff report a problem, raise a support ticket and check its progress without leaving the chat.
  • It only reads what you are allowed to read. Permissions are applied at the moment documents are fetched, using the company existing Azure AD groups, so nothing you are not cleared for ever reaches the answer.
  • Every release is tested for two things before it ships: is the answer correct, and did it respect access rules. Failing either one blocks the release.
Read the detail
Architecture diagram for A company assistant 20,000 people actually useAzure ADEmployee. Select to see why it is here.EmployeeAssistant. Select to see why it is here.AssistantFind documents. Select to see why it is here.Find documentsDocuments. Select to see why it is here.DocumentsActions. Select to see why it is here.ActionsIncidents and tickets. Select to see why it is here.Incidents andticketsEvaluation framework gates every release

Click any box in the diagram to see why it is there

02CreditRiskMonitorIn pilot

Letting customers ask an AI about credit data, safely

Customers point their own AI assistant, such as ChatGPT or Claude, straight at our credit risk data. Every request is checked against what that customer actually pays for.

  • Built on the Model Context Protocol, an open standard that lets an AI assistant call a company data directly instead of a person logging into a portal.
  • Three checks on every single request: who is asking, what they are entitled to, and a permanent record of what was sent to them.
  • Each customer is walled off from every other. One customer data can never surface for another, and no single customer can slow the service down for everyone else.
Read the detail
Architecture diagram for Letting customers ask an AI about credit data, safelyTool callAI client. Select to see why it is here.AI clientMCP server. Select to see why it is here.MCP serverWho are you, what did you buy. Select to see why it is here.Who are you, whatdid you buyScores and peers. Select to see why it is here.Scores and peersKept separate, kept fair. Select to see why it is here.Kept separate,kept fairEvery call is written to an audit log

Click any box in the diagram to see why it is there

03CreditRiskMonitor

Safe guardrails for 50 developers using AI

Developers use AI to help write code. This sets the boundary of what those AI tools are allowed to touch, the same way for everyone.

  • One shared setup instead of 50 people each deciding for themselves what an AI tool may access.
  • Spells out exactly which tools, which code and which company data an AI assistant can reach, and blocks the rest.
Read the detail
Architecture diagram for Safe guardrails for 50 developers using AIDeveloper. Select to see why it is here.DeveloperThe guardrails. Select to see why it is here.The guardrailsAllowed tools. Select to see why it is here.Allowed toolsAllowed code. Select to see why it is here.Allowed codeAllowed data. Select to see why it is here.Allowed dataThe AI tool. Select to see why it is here.The AI toolThe harness decides the reach, not the agent

Click any box in the diagram to see why it is there

Built with

  • Python
  • JavaScript
  • Node.js
  • C#
  • .NET
  • React
  • Next.js
  • SQL
  • PostgreSQL
  • Redis
  • Docker
  • Azure
  • Azure OpenAI
  • MCP
  • RAG
  • Agentic workflows

Hardware and IoT

Not everything I build runs in a data centre.Some of it sits in a factory and cuts the power bill.

I have worked on energy since 2016. At CodeElectra we built a device that measured a building electricity use and let you control it remotely. That is where the hardware started.

I co-founded Varnilix Private Limited in May 2025, an industrial technology company recognised by DPIIT, the Indian government startup scheme. It builds V-GRID, which fixes a hidden waste in factory electricity: factories are billed for power they never actually use. V-GRID cuts that waste, so the bill goes down without changing a single machine.

I also lead the software products built at Varnilix, including a multi tenant workforce and payroll platform and a web based HR management system.

DPIIT recognisedIndustrial IoTPower electronics

Patent applications

Both cover the same idea in different ways: how to cut that wasted electricity automatically, and how to do it without the equipment ever becoming unsafe.

01System and methods for distributed autonomous reactive power compensationFirst examination report 12 May 2026, response filed August 2026202621022494
02Power factor correction apparatus with deterministic hardware safety interlocks and relay weld detectionExpedited examination, first examination report issued202621016587

Named inventor on two Indian patent applications, filed in February 2026 and under examination. Varnilix Private Limited is the applicant.

  • RecognitionEmergency Vehicle Navigator, an IoT ideaIntel Higher Education, shortlisted in a national selection, 2017
  • LeadingStudent head for a national level hackathon1,800 participants on campus, leading 150 volunteers

Timeline

Ten years of shipping, in short.

  1. 2015First paid workWashim District PoliceBuilt the new website for the Cyber Cell at the Superintendent of Police office, while still studying. Recognised with a commendation letter that December.
  2. 2017Co-founder, director and IT specialistCodeElectra TechnologiesStarted setting the company up in October 2016 and incorporated it in 2017, then ran it to August 2021. Our own product was a wireless electricity consumption monitoring and control system over IoT. Incubated at the Savitribai Phule Pune University incubation centre in 2019, and mentored around 150 students along the way.
  3. 2020B.E. Computer EngineeringSavitribai Phule Pune UniversityGraduated First Class from G. H. Raisoni Institute of Engineering and Technology.
  4. 2021Associate, Education, Training and AssessmentInfosysTrained and mentored more than 500 engineers worldwide, and built the online assessment system used across the programme.
  5. 2023Senior Software EngineerBroseIncident reporting and automation for the security operations and IT service management teams across Brose sites. On site in Bamberg, Coburg and Wuerzburg.
  6. 2024Technical LeadBroseLed four engineers building the internal generative AI assistant.
  7. 2025Co-founder and directorVarnilix Private LimitedCo-founded a DPIIT recognised industrial IoT company building V-GRID.
  8. 2026Engineering LeadCreditRiskMonitorLeading the India engineering team of 15 at a NASDAQ listed credit risk company, engaged through Encore Infotech Services.

Contact

Let's build something that ships.

Ready to relocate.

akshaykankal@gmail.comLinkedIn
Akshay Kankal

Languages

English C1 · German A1, learning · Marathi native · Hindi native