● Data Scientist & Analytics Leader

I build data science teams that answer to outcomes, not outputs.

Fourteen years across energy, banking, retail and CPG in the US and India. I joined a bank as an individual contributor and left with a 21-person team and 20+ models running in its CRM decisioning layer. Today I detect meter bypass in AMI interval data at a Fortune 500 utility.

See the impact Résumé LinkedIn
  • $60M+
    in new savings deposits from a Voice BOT channel I launched
  • $1.8M+
    in savings, recovered leakage and incremental revenue
  • 0→21
    data scientists hired and led at RBL Bank, including 5 team leads
  • 20+
    ML models wired into a near-real-time CRM decisioning layer
01 — Impact

The dollars, itemised.

Every figure here is on my résumé and traces to a specific piece of work.

$1.8M+savings, recovered leakage and incremental revenue
  • Revenue leakage recoveredRBL Bank · lien recovery & debit-card charging$850K+
  • Inventory savings, per yearApollo Pharmacy · forecasting & simulation$600K+
  • Incremental revenue from 8+ modelsRBL Bank · XGBoost, Random Forest, CF$360K+
  • Purchasing savings, per yearTasty Bite · commodity price forecast$12K+
$60M+ in new savings-account deposits from a Voice BOT acquisition channel I launched IDFC First Bank · counted separately, because deposits aren't revenue
  • 33% vs 20%energy-diversion hit rate against field methodsDominion Energy · pilot
  • ↓25%customer acquisition cost through A/B-tested campaignsIDFC First Bank
  • ↑25%cross-sell revenue from second-product recommendationsIDFC First Bank
02 — Career

Fourteen years, one through-line.

From automating builds at TCS to leading data science teams in banking and retail. Most recent first.

Mar 2025 – now Senior Data Science Analyst (contract) Dominion EnergyFortune 500 utility · Richmond, VA
Contract role
  • Designed an energy-diversion detection solution that combines rule-based logic with anomaly detection on AMI interval data. In the pilot, model-surfaced accounts confirm diversion at a 33% hit rate, against 20% for existing field methods.
  • Built enrollment propensity models for eBill, AutoPay and Budget Billing across 2.8M customers, enabling targeted outreach projected to deliver a 10% relative lift in enrollment.
  • Deployed a payment-delay prediction model for proactive collections, improving early identification of at-risk accounts by 50%.
  • Built an EMS AI comparative-analytics platform and interface that ranks event-related environmental risk, using clustering, predictive modelling and similarity analysis.
Dec 2023 – Dec 2024 Full-time graduate study University of ArizonaM.S. Information Science · 4.0 GPA
Dean's List
  • Focused on machine learning, NLP and deep learning, and was named to the Dean's List of Distinguished Scholars.
  • Capstone: SoulGuard, a deep-learning and LLM app that flags depression-induced suicide risk.
  • Group project: Bayesian regression and MCMC on global commodity emissions.
Sep 2022 – Dec 2023 Vice President, Lead Campaign Planning IDFC First Bank$40B+ assets · Mumbai
Team of 5
  • Led a 5-analyst campaign analytics team owning measurement and targeting across the retail marketing portfolio.
  • Established A/B testing and causal inference as the standard for campaign sign-off, replacing pre/post reporting and cutting customer acquisition cost by 25%.
  • Built ML-driven second-product recommendations that prompted call-centre agents to attach another product after the first sale closed, lifting cross-sell revenue by 25%.
  • Launched a Voice BOT acquisition channel as a new digital origination path, generating $60M+ in new savings deposits.
  • Raised customer product holding by 8%. ACE Award for leadership in BI, 2023.
May 2019 – Sep 2022 Deputy Vice President, Analytics Lead RBL Bank$20B+ assets · Mumbai
Team of 21 · 5 leads
  • Built and led a 21-member data science team, including 5 direct-report team leads, delivering analytical use cases across the bank.
  • Architected an AWS data pipeline integrating 20+ ML models into a near-real-time CRM decisioning layer, lifting cross-sell revenue by 10%.
  • Delivered 8+ production models (XGBoost, Random Forest, collaborative filtering) generating $360K+ in incremental revenue.
  • Recovered $850K+ of revenue leakage through data mining across lien recovery and debit-card charging workflows.
  • Developed a customer-lifecycle (CLCM) framework for liability and asset products using external data, including bureau information, raising customer retention by 10%.
  • Designed multi-channel interaction dashboards, including a Voice-of-Customer view across email, chat and calls, that improved customer experience and reduced operational cost.
Apr 2016 – May 2019 Senior Manager, Inventory Analytics Apollo Pharmacy5,000+ stores · Hyderabad
Team of 5 · built from scratch
  • Led a team of 5 data scientists owning inventory optimisation across 3,000+ stores and 8 distribution centres.
  • Deployed demand forecasting and simulation models that reduced stockouts by 20% and excess inventory by 25%, delivering $600K+ in annual savings.
  • Improved product availability from 87% to 91%. Geospatial analysis of hub-store hotspots optimised DC-to-store routing and cut emergency delivery time by 3 hours.
  • Ran causal studies combining transactional and survey data, which drove changes to store operations and training programmes.
  • Designed the analytics framework for the Physical Bounce Audit and a scalable data-lake setup for the team.
May 2013 – Apr 2016 Assistant Manager Tasty Bite EatablesConsumer packaged goods · Pune
Cross-functional teams
  • Built competitive-intelligence dashboards on U.S. SPINS data to compare products shelf by shelf across retailers.
  • Built a commodity price forecast for brown-onion paste that saved $12K+ a year in purchasing.
  • Raised plan compliance from 80% to 95%, production compliance from 95% to 98% and shipment OTIF from 80% to 92%.
  • A time-and-motion study raised line throughput by 20%. I coached a process team that cut water use by 50,000 litres a day.
Jul 2008 – Apr 2012 Module Lead Tata Consultancy ServicesGlobal IT services · Hyderabad
Ericsson OSS plugins
  • Led the BBOSS configuration team and owned end-to-end delivery.
  • Automated build and release, cutting build time by 75%, and added scheduled builds as a quality gate.
  • Won two On-the-Spot awards: one for build automation and one for 100+ deliveries in a single month.
03 — Projects

Still hands-on after hours.

Leading teams is the day job. These are things I build myself. All of them are still being worked on, so expect rough edges and frequent changes.

Open source · PyPIpip install

py-flow-mapper

A beta Python package that turns a codebase into Mermaid flow diagrams using static AST analysis. It shows which functions call which and how return values flow between them, without running the code or depending on an LLM.

# map any project in two commands
$ pip install py-flow-mapper
$ pyflow analyze ./my_project
$ pyflow diagram project_meta.json
  • Python 3.12
  • AST
  • Mermaid
  • GitHub Actions → PyPI
github.com/ArunKoundinya/py-flow-mapper →
Personal AI assistantPrivate

ArunJarvis

A voice assistant with its own brain and memory, plus interchangeable "ears" and "mouths". The brain is a plain text-in, text-out service, so Alexa is just the first surface. A phone camera page gives it "eyes": it compares frames over time, so it can answer what has changed.

  • Alexa Skills
  • Supabase Edge Functions
  • Deno / TypeScript
  • Postgres memory
  • LLM APIs
Code available on request
Capstone · U of ArizonaLive site

SoulGuard

A suicide-prevention app that uses deep learning and NLP on social-media posts and user input. It detects signs of depression and tells general symptoms apart from suicide risk, with two modes: "For You" and "For Whom You Care".

  • Deep learning
  • NLP
  • LLMs
  • Python
  • CI testing
arunkoundinya.github.io/SoulGuard →
Full-stack PWAPrivate repo

Fitness analytics platform

A multi-user progressive web app for workout logging and nutrition tracking. Each set shows last session's number next to today's input, and an insights layer shows progression and adherence trends.

  • React + Vite
  • Supabase Postgres
  • Row-level security
  • pgTAP
  • Vercel
Code available on request
LLM persona enginePrivate · live

Sthayi

An experiment in characters that change over time. Most chatbots treat a persona as a fixed prompt. In Sthayi, the relationship writes to a state, and that state shapes how the character speaks each turn: register, directness and willingness to push back. It has seven memory threads and group conversations, and is now in a month of daily use to see whether week four feels different from week one.

  • Next.js
  • Supabase Postgres
  • Embeddings
  • LLM APIs
  • Vercel
Code available on request
Automation

Job-search automation

A tool that helps recent graduates automate job searches across sites. It uses a Selenium scraper with a Django app to store and review the results.

  • Selenium
  • Django
  • Codespaces
View repo →
04 — Writing & teaching

Thinking out loud, in public.

Essays since 2017, a YouTube series on AI basics, weekly data-viz practice, and regular LinkedIn commentary on how analytics teams should be organised and held accountable.

Canvass And Analyze

Essays · since 2017

Personal essays and long-running series where psychology, Indian culture, analytics and management meet.

Read the essays →

AI Basics with AK

Video series · notes & code

A YouTube vlog series on AI fundamentals, with notes, code and companion blogs for each season.

Watch & learn →

Tidy Tuesday

R & Python · since 2024

My solutions to the weekly TidyTuesday community challenge, from first line plots to fuller analyses.

See the charts →

Deep Learning notebook

TensorFlow · study notes

Deep-learning notes, published as a website so I can review them on my phone.

Open the notes →
Toolkit

What I reach for.

Grouped by what each tool is for. The case studies above show how I've used them.

Modelling

  • Forecasting
  • Supervised & unsupervised ML
  • Anomaly detection
  • Optimisation & simulation
  • NLP

Experimentation

  • A/B testing
  • Causal inference
  • Design of experiments
  • KPI design

LLMs & retrieval

  • Agentic systems
  • RAG patterns
  • Vector search (FAISS)
  • Prompt design
  • Local LLMs

Languages

  • Python
  • R
  • SQL
  • Shell

Platforms

  • AWS
  • Databricks
  • Snowflake
  • Dataiku
  • Azure DevOps
  • Oracle Exadata
  • GitHub

Communicating

  • Power BI
  • Tableau
  • ThoughtSpot
  • Executive storytelling

Leading

  • Hiring & building teams
  • Team mentorship
  • Use-case design
  • Stakeholder engagement
  • Vendor negotiation
  • Change management

Education

  • M.S. Information Science (Machine Learning)University of Arizona · Tucson, AZGPA 4.0/4 · 2024
  • Certificate in Business AnalyticsIndian School of Business · HyderabadGPA 3.6/4 · 2015Graduated with a merit certificate while working full-time.
  • Post Graduate Programme in ManagementGreat Lakes Institute of Management · Chennai7.9/10 · 2013Founded the Economics Club and launched Arthayuga, a national-level college event.
  • B.Tech, Electrical & Electronics EngineeringVellore Institute of Technology7.9/10 · 2008
  • IBM Data Science Professional
  • Deep Learning Specialization
  • MLOps Specialization
  • Databricks Fundamentals

Recognition

  • Dean's List of Distinguished ScholarsUniversity of Arizona2024
  • ACE Award, leadership in Business IntelligenceIDFC First Bank2023
  • Dean's List, top 10% of the cohortIndian School of Business2015
  • Gold Medal for leadershipGreat Lakes, for founding and running the Economics Club2013
  • On-the-Spot Award ×2Tata Consultancy Services, for build automation2011
05 — Contact

Building an analytics team that has to earn its keep?

I'm glad to talk about data science leadership roles, hard modelling problems, or how to make analytics accountable to business outcomes.

Emailparasa.arunkoundinya@gmail.com LinkedIn/in/arunkoundinya0710 GitHub@ArunKoundinya RésuméWord document