Madhu Sudhan Srinivas

Dublin 9, Ireland · smadhu6364@gmail.com · github.com/smadhu6364-beep · linkedin.com/in/madhu-sudhan-a966b3225

I've run day-to-day operations on busy front lines, and now I build data and AI tools. I'm looking for data and analytics roles in Ireland.

Education

MSc in Business Analytics (Computing)

2026 – present

Dublin City University, Dublin

Modules: Statistics; Visualisation & Communication (Power BI); Programming & Machine Learning (Python); Research Skills

BE Computer Science Engineering

2019 – 2023

DSATM, Bangalore

Experience

Operations Executive (Customer Service), Thomas Cook (India) Limited

Apr 2026 – Sep 2026

Bengaluru

  • Audited booking records for errors and corrected them, keeping customer data accurate.
  • Resolved a high volume of booking queries and issues, working each one back to the record that caused it.
  • Worked with the operations team to keep daily processes and records running smoothly.

Operations Coordinator, Trip Factory

Jul 2024 – Mar 2026

Bengaluru

  • Compiled daily shift figures into Excel tracking sheets that supervisors used to plan the week.
  • Ran stock checks and inventory reconciliation at set checkpoints.
  • Handled customer onboarding and card payment processing, and resolved complaints in a high-footfall setting.

Cafe Team Member, Amitri Cafe

6 months

Bengaluru

  • Took orders and kept service moving at peak times.
  • Kept to food hygiene standards.

Projects

LLM Risk Register Benchmark · github.com/smadhu6364-beep/public-reproducible-benchmark

A benchmark study that evaluates three LLMs across three prompting strategies for automated risk-register generation. Inputs are real project planning documents from 21 World Bank and UK government sources. Reproducible pipeline with a strict no-leakage evaluation design, covered by 456 unit tests. Being prepared for an IEEE conference.

Python, LLM evaluation, Prompt engineering, Reproducible research

Irish property prices · github.com/smadhu6364-beep/irish-property-sql · live demo

Every residential sale on the Property Price Register from 2010 to October 2026, analysed in nine SQL files. The national median was €380,000 in 2025, up from €148,000 in 2013. Dublin's lead over the rest of the country has shrunk from 2.22 times to 1.49 times.

SQL, DuckDB, Python, pandas, matplotlib

Irish property dashboard · github.com/smadhu6364-beep/irish-property-sql · live demo

KPI tiles, a county filter, a year range, a price trend against the national median and a county league table. Built on a new window-function query (LAG for year-on-year change, RANK by county) over the property results.

SQL, DuckDB, React, TypeScript

CivicPulse · github.com/smadhu6364-beep/civicpulse

Take a photo of a local problem. Gemini Vision categorises it, scores its severity, works out which authority is responsible and drafts a complaint letter. Reports appear on a live community map.

React, FastAPI, Gemini Vision, Firebase, Leaflet

Roam · github.com/smadhu6364-beep/roam-travel · live demo

Front-end prototype for planning and booking trips. Flights and payments are scaffolded, not live.

Next.js 16, React 19, TypeScript, Tailwind

In-browser vision demos · github.com/smadhu6364-beep/aiml-project · live demo

A set of demos using BodyPix, Face Mesh, Handpose and COCO-SSD with TensorFlow.js and React. Everything runs client-side, no server needed.

TensorFlow.js, React, BodyPix, Face Mesh, Handpose, COCO-SSD

Skills

Data
Python, SQL, DuckDB, pandas, Advanced Excel, Power BI, Statistics, Data visualisation
Web
JavaScript, TypeScript, React, Next.js, FastAPI, Leaflet.js, Firebase
AI / ML
Gemini API, Prompt engineering, TensorFlow.js, LLM evaluation, Reproducible research
Tools
Docker, Git, GitHub Actions
Operations
Customer service, Query and complaint resolution, Records accuracy, Inventory checks, Shift reporting