Senior Data Scientist at Atlassian · Bengaluru

Aditya Yadav

Most of my work is building ML systems and, lately, AI agents, and then checking carefully whether they actually help. On the side I'm building VettaCV, a small app for people looking for their next job.

Book a chat aditya.yadav.of@gmail.com

LinkedIn GitHub Resume Open to relocation

Currently at Atlassian

8 yrsworking with data

VettaCVmy side project

How I tend to work

01 — Work

What I spend my time on.

01 · ML SYSTEMS

Models that run every day, not just in a notebook.

The data, the features, serving it, and keeping an eye on it once it's live.

scikit-learnPyTorchPySparkSageMaker

At Grab, fraud models I worked on brought the incident rate down 30 points.

02 · AI AGENTS

Agents, and the unglamorous part: knowing when they're wrong.

Building with LangGraph and Claude, then measuring frustration, errors, latency and cost per conversation.

LangGraphClaudeRAGevals

A support agent I worked on at Atlassian now handles 50% more questions on its own.

03 · CONSULTING

The fuzzy questions

Sitting with a team that has a vague question and turning it into something we can measure and act on.

Fed into safety changes that cut incidents in Thailand by 13%.

04 · EXPERIMENTS

Experiments & causal inference

A/B and switchback tests, power analysis, causal methods. Mostly in service of: did it actually work?

One confirmed a 25-point jump in harmful-content detection.

05 · AUTOMATION

Pipelines that look after themselves

Replacing fragile scripts and manual steps with tested, scheduled pipelines.

~40 legacy SQL jobs moved to dbt, three months ahead of plan.

02 — Tools

What I usually reach for.

ML & deep learning

scikit-learnPyTorchTensorFlowBERT & transformersNLP & text classificationAnomaly detectionForecastingRecommendations

Agents & LLMs

LangGraphLangChainClaudeOpenAI APIStructured outputsRAGAgent evals & observabilityPrompt & token optimisation

Data & pipelines

PythonSQLPySparkdbtDatabricks

Experimentation

A/B testingSwitchback experimentsPower analysisHypothesis testingCausal inference

Dashboards & BI

Power BITableauGrafanaMixpanelDatabricks dashboards

Cloud & apps

AWS · SageMakerGCPFirebaseFlaskNext.jsGit

03 — Career

Where I've been.

Atlassian · Bengaluru

2025 – now

Senior Data Scientist · Cloud migration

Looked after

Analytics and experiments for moving customers from Data Center to Cloud, from deciding what to measure to walking leadership through it each week.

Built

Experiments for an AI migration-support agentAgent observability metricsCustomer complexity classifierDBT rebuild of ~40 legacy SQL jobs

What changed

+50%more questions handled by the migration-support agent

−20%time to resolve a migration issue

24 → 21days for a typical customer migration

Grab · Southeast Asia

2022 – 2024

Data Scientist · Trust, Integrity & Safety

Looked after

Started content-moderation analytics from scratch, then looked after fraud and safety analytics for an app used by 10M+ people.

Built

Fraud models: collusion, impersonationRisk-based user verificationUser Friction ScoreA/B & switchback experimentsC-level safety readouts

What changed

13%fewer safety incidents in Thailand

30pplower fraud rate, about SGD 200K a month in bookings kept safe

70ppless friction in verification, still catching 80% of fraud

J.P. Morgan · Mumbai

2021 – 2022

Senior Analyst · Thematic indices

Looked after

Thematic equity indices, from the NLP signal underneath to how the index behaved once it was running.

Built

BERT sentiment pipeline on newsTheme scoring for AI & renewablesVolatility-based index

What changed

A fewthematic indices, AI and renewables among them

$100Kin client value

Netradyne · Bengaluru

2020 – 2021

Data Analyst · Driver-safety telematics

Looked after

Analytics on live telemetry from vehicles, aimed at helping drivers stay safe.

Built

Pipeline automationTelemetry anomaly detectionAccident-hotspot clusteringSQL → NoSQL migration

What changed

1/6thruntime cut by automating the pipeline

12%fewer accidents in flagged zones

IndShine · Delhi

2017 – 2020

Senior Data Analyst · Drones & geospatial

Looked after

Pricing and mapping analytics for drone surveys in construction, agriculture and government projects.

Built

3D maps from 2D imageryTimeline prediction modelsSegment-based pricing

What changed

20%more accurate project timelines

5%revenue uplift from pricing

04 — Moments

Outside the notebook.

The people and places around the work: team dinners, offsites, and the odd pickleball game.

Scroll sideways · click a photo to enlarge

05 — Side projects

Things I build on the side.

VettaCV · side project

A small app that helps people land their next job.

It's a one-person project these days, so I end up touching everything from the LLM pipeline to the payment page.

Rewrites your resume for a specific job, without making anything up
Finds openings that fit and helps with the application questions
Lets you practise interviews, with AI or with people who've done the job

About 400 people used it this year, from seven countries, and that number has been climbing.

Visit vettacv.com → Notes on building it

SnapMyDesign · AI photoshoot generator

Product photos without a photographer.

An earlier experiment: text-to-image and style-transfer models wrapped in a simple flow, so non-designers can make product shots and try-ons themselves.

Visit snapmydesign.com → Details

Case studies & experiments

WerkspotDemand–supply gaps and a switchback test for a new matching algorithm.Case study→ BoltA data-backed acquisition strategy for a mobility startup.Case study→ AgodaBusiness opportunities in hotel pricing trends.Case study→ Book recommenderA recommendation system for books.ML→

06 — Say hello

Happy to talk.

If any of this overlaps with what you're working on, I'd like to hear about it. Away from work I've taught with Teach for India, and these days it's mostly tennis and swimming.

Book a chat Email LinkedIn