Skip to content
Mirun Kaushik

AI backend engineer / India

MirunKaushik

I build the backend that AI products run on. Rules first, and a model only where it earns its place.

Now
AI Engineer Intern
Botcadence
Before
Full-Stack AI Engineer Intern
Euredas.ai
Scroll
01Approach

Rules first.The model onlywhere it earns its place.

I structure data with rules first and call a model only where it is needed. It costs less, hallucinates less, and the output is easier to predict.

Before anything reaches a model, I structure it with ontology rules: what a medical report says, or which claim on a web page is worth checking. What’s left is small and specific, and that is the only part a model sees.

I used it in Health Link and Mind Armor.

A model at every step

Rules first

Fig. 2 — The same pipeline, two ways. A schematic, not a benchmark.Fewer calls · fewer surprises · cheaper to run
02Path

From hackathon stagesto production code.

Three podium finishes in two years. The second place at Hackwell turned into an internship offer from Euredas.ai, and from there I moved to production work at Botcadence.

03Selected work

Things I built,and who built the rest.

Each project says what I built, what my team built, and what is still only a design.

04Experience

Experience

Aug 2026 – Present

Botcadence

AI Engineer Intern

Multilingual AI voice-agent platform · Remote

A subscription platform where clients sync contacts from their CRM and run outbound campaigns: voice-agent calls in the customer’s own language, plus SMS and WhatsApp. The agent collects information during the call and writes it back to the CRM. I work with a US-based team across time zones.

What I built

  1. 01Integrated Zoho CRM, Salesforce and Pipedrive so clients can sync contacts into campaigns and have call outcomes written back to their CRM.
  2. 02Built post-call webhooks that send call details and AI-generated insights as JSON to any endpoint the client configures.
  3. 03Implemented role-based access control across the application and tested each feature end to end before release.

How a campaign moves through it

  1. 01Client CRMZoho · Salesforce · Pipedrive
  2. 02Contact syncinto campaigns
  3. 03Voice agent callnative language(model call)Team
  4. 04Write-backoutcomes to CRM
  5. 05Post-call webhookJSON to client endpoint
Model call Rules, data, queues Built by teammates

Mar 2026 – Aug 2026

Euredas.ai

Full-Stack AI Engineer Intern

Loyalty-points CRM for local businesses · Tiruchirappalli

A loyalty-points CRM launching for local businesses in Trichy. I joined after placing 2nd at Hackwell 2026.

What I built

  1. 01Built a brand-asset extractor: give it a website URL and it parses the HTML, CSS and DOM to pull out the brand’s colours, logos, fonts and tagline.
  2. 02Implemented product web pages and features from product specs, testing each flow end to end.
05About

I decide thearchitecture,check whatthe tools write,and test it end to end.

I’m a final-year AI and Data Science student in Tiruchirappalli, working as an AI engineer intern at Botcadence. My work there is the plumbing behind a voice-agent product: getting contacts in from a client’s CRM, getting call outcomes back out, and deciding who is allowed to see what.

Most of what I know came from hackathons. Three podiums in two years taught me to scope hard, build the part that matters, and be honest about what is still a slide.

I build with AI coding tools every day. I decide the architecture and the data flow, check what the tool writes, and test every feature end to end before it ships.

Looking forAI and backend engineering roles at product companies and startups in India.

Education

8.66CGPA / 10

B.Tech Artificial Intelligence and Data Science (Honours)

Saranathan College of Engineering, Anna University

Aug 2023 – Mar 2027 · Tiruchirappalli, Tamil Nadu

Certified

  • Microsoft Azure FundamentalsAZ-900
  • Microsoft Azure AI FundamentalsAI-900

Toolkit

Languages
Python, JavaScript, TypeScript, SQL
Backend
Node.js, REST APIs, Webhooks, Role-based access control
Data & queues
MongoDB, Redis, pgvector, BullMQ, Celery
AI
RAG, LLM agents, Vector search, Hugging Face
Tools
Git, Claude Code, Cursor, Antigravity
06Contact

Based in

Tiruchirappalli, India

--:-- IST