Skip to content

Data Scientist · AI/ML Engineer · Perth, Australia

Hi, I'm Raja Venkata Shandilya Puram. Call me Shandy.

I build AI that works in the real world.

Machine learning, computer vision, LLM apps and cloud data pipelines, validated against ground truth.

See my work
Resume
LinkedIn
GitHub

Open to relocate nationally and ready to work remote.

Scroll
Mission 02

Telemetry

01
production pipelines shipped
02
video frames processed in one run
03
better than Random Forest & HMM baselines on unseen data
04
faster queries over 15 GB+
05
uptime for 50+ daily users
06
policy institutions adopted his dashboards
Mission 03The pilot

Messy data in. Trusted decisions out.

I'm Shandilya. I turn raw, messy data into automated, validated pipelines that save experts days of work.

That has meant geological intelligence for a government agency, lab video analysis at the Perron Institute, and natural-language-to-SQL at Hyperspace.

I test every result against ground truth, because a model only matters if experts can trust it.

Machine Learning → Scikit-Learn · Random Forest · HMM · TensorFlow
Mission 04Featured missions

Three systems. Real users. Measured results.

UWA Capstone · Geological Survey of Western Australia

GSWA HyLogger Industry Adoption

Team Lead & ML/AI Lead

Led a 6-person team building an automated geological intelligence pipeline on hyperspectral drill-core data.

  • nvcl-kit retrieval → mineral percentages → rule-based classifier
  • PostgreSQL/PostGIS, FastAPI backend, React explorer, Docker Compose
  • Owns ML/AI: baseline classifier delivered; building anomaly scoring, spectral similarity search and deep-learning property prediction
person team
person team
commits, PR-reviewed
commits, PR-reviewed
  • Python
  • PostGIS
  • FastAPI
  • React
  • Docker
View codeLive explorer
Drill core · hover an intervalExample data
120 mdepth →148 m
Interval
132–134 m
Flag
anomaly
Dominant mineralCarbonate
Kaolinite8%
Muscovite20%
Chlorite22%
Carbonate50%

Perron Institute · Research Data Scientist Intern

Automated Behavioural-Analysis Pipelines

Research Data Scientist (Intern)

Five production pipelines that turn DeepLabCut pose output into validated HTML and Excel reports.

  • ~13% better generalisation than Random Forest and HMM baselines
  • 60–70% agreement with manual ground truth
  • Found and fixed a silent tracking bug

Prof. Jennifer Rodger said the tools substantially increased her team's capacity.

production pipelines
production pipelines
agreement with ground truth
60–70%
agreement with ground truth
  • Python
  • DeepLabCut
  • Pandas
  • Scikit-Learn
Read the story
Pose → validated reportIllustration
keypoints · nose / ears / tail-base
Generalisation on unseen data
RF / HMMbase
Pipeline~13%↑

Hyperspace · AI/ML Engineering Intern · Aug 2024 – Jan 2025

Natural-Language-to-SQL Analytics

AI/ML Engineering Intern

Clients gave raw operational data and had no easy way to ask it questions.

Shandilya built an end-to-end pipeline where anyone types a question in plain English and gets a clear written answer.

faster query response
faster query response
uptime for 50+ daily users
uptime for 50+ daily users
sellers get automated written insights
sellers get automated written insights
  • FastAPI
  • SQLAlchemy
  • PostgreSQL
  • OpenAI API
  • Python
GitHub
Pipeline · 6 steps

01 · Cleaned and loaded into PostgreSQL (15 GB+ of structured data)

Live demoIllustration with sample data, not the production system

› Which sellers grew fastest last month?

WITH m AS (
  SELECT seller_id, date_trunc('month', created_at) AS month, SUM(amount) AS revenue
  FROM orders GROUP BY seller_id, month
)
SELECT s.name, (cur.revenue - prev.revenue) / prev.revenue AS growth
FROM m cur
JOIN m prev ON prev.seller_id = cur.seller_id AND prev.month = cur.month - interval '1 month'
JOIN sellers s ON s.id = cur.seller_id
WHERE cur.month = date_trunc('month', now()) - interval '1 month'
ORDER BY growth DESC LIMIT 3;

answerSeller 14 grew fastest last month, with revenue up 32% on the month before.

Mission 05Side missions

Side missions

  • Graph05.01

    Aviation Graph Intelligence

    57,301 routes in a graph, with a Codeshare Gap Detector and a Ghost Bridge Airport analyser.

    • Neo4j AuraDB
    • Cypher
    • APOC
    View code
  • ML05.02

    BoxOffice Intelligence

    Four models reaching AUC 0.85, with explainable predictions.

    • R
    • Shiny
    • caret
    • LIME
    • DALEX
    View code
  • Data Engineering · ML05.03

    London Bus Safety Analysis

    Full ETL on AWS, plus a Random Forest with 85% accuracy.

    • AWS S3
    • Glue
    • Python
    • Tableau
    View code
  • Analytics05.04

    99 Bikes & Global Financial Inclusion

    Dashboards over 20,000+ transactions and World Bank Findex data.

    • Power BI
    • Tableau
    View code
  • Analytics05.05

    NASA Global Landslides

    Exploratory analysis of NASA's global landslide data.

    • pandas
    • NumPy
    • Matplotlib
    View code
  • Data Engineering05.06

    PillPal

    A Python app made with Streamlit.

    • Python
    • Streamlit
    View code
  • ML05.07

    BCG X Data Science

    EDA and modelling in Python for the BCG X data science programme.

    • Python
    • EDA
    • Modelling
    View code
  • Analytics · ML05.08

    Student Outcomes Dashboard

    Regression on 15,000+ records, adopted by 3 policy institutions.

    • Power BI
    • Regression
    View code
  • Analytics05.09

    TfL Accidents Dashboard

    Tableau dashboard of Transport for London accident data.

    • Tableau
    View code
Mission 06Flight log

Where I've flown.

Oldest at the top, latest at the bottom.

  1. Jun – Aug 2024

    Data Analyst Intern

    Cloud Counselage · Remote

    Regression on 15,000+ records found a 22% competence gap; Power BI dashboards adopted by 3 policy institutions.

  2. Aug 2024 – Jan 2025

    AI/ML Engineering Intern

    Hyperspace · Hyderabad

    Plain-English questions to SQL to written answers: ~40% faster query response, 99.8% uptime for 50+ daily users.

  3. Jun – Jul 2026

    Research Data Scientist (Intern)

    Perron Institute / McCusker Centre

    Five production behavioural-analysis pipelines from DeepLabCut output to validated reports.

  4. Semester 2, 2026

    Project Team Lead & ML/AI Lead

    GSWA · UWA Capstone

    Led a 6-person team on an automated geological intelligence pipeline.

Mission 07Toolkit

Instruments on the panel.

Languages

09 tools
  • Python
  • SQL
  • R
  • PySpark
  • NumPy
  • Pandas
  • Scikit-Learn
  • OpenCV
  • FastAPI

ML & AI

09 tools
  • Classification
  • Regression
  • Random Forest
  • Hidden Markov Models
  • TensorFlow
  • OpenAI API
  • LIME
  • DALEX
  • Computer Vision

Data & Cloud

11 tools
  • AWS S3
  • Glue
  • SageMaker
  • Docker
  • Git
  • Azure DevOps
  • PostgreSQL
  • PostGIS
  • Neo4j
  • Cypher
  • ETL

BI

04 tools
  • Power BI
  • Tableau
  • R Shiny
  • Excel
Mission 08Credentials

Trained, certified, involved.

Education

  • Master of Data Science

    University of Western Australia

    Feb 2025 – Dec 2026

    International Scholarship ($6,000)

  • B.Tech Computer Science & Engineering

    JNTU Hyderabad

    2020 – 2024

Certifications

  • Forage · Quantium Data Analytics
  • Forage · BCG X Data Science
  • IBM · DS0101EN
  • HackerRank · SQL
  • Cisco · Python Essentials
  • Udemy · Python Bootcamp

Community

  • UWA Open Day volunteer, guided 100+ students
  • Volunteer Power BI analyst for 4 retail branches
  • UWA Data Science Club
Mission 09Open a channel

Got data that eats your team's time? Let's talk.

Personal

rajavenkatashandilyap@gmail.com

UWA

24944809@student.uwa.edu.au
LinkedIn
GitHub
Resume

Hi! Scroll. It gets better.