Sai Nimisha Kakarla
Menu

Data analytics portfolio

Sai Nimisha Kakarla

Senior Data Analyst · SQL, Python & Tableau

I turn complex, multi-source data into governed KPIs and trusted insights that drive business decisions.

View projects Resume Get in touch
01

About

I build end-to-end analytics: warehouse models, SQL transformation layers, automated data-quality checks and self-serve Tableau dashboards that give business teams a single trusted view of their numbers.

I care about numbers people can sign off on: I partner with finance, merchandising and marketing to define KPIs, validate every one against source data through UAT, and document schemas, lineage and logic for easy handoff.

Discipline
Data analytics
Current employer
Neiman Marcus Groups
Domains
Retail, healthcare claims, supply chain
Degree
MTech Integrated Software Engineering, VIT University, Chennai (June 2015 – June 2020)
02

Experience

Neiman Marcus Groups

Data Analyst

Apr 2023 – Present

Dallas, TX

Owns end-to-end retail analytics across sales, returns and inventory, from the PostgreSQL warehouse and SQL transformation layer to KPI definitions and self-serve Tableau dashboards.

Highlights
  • Architected a retail analytics solution unifying 7 sources (OMS, POS, product master, marketplace, marketing, customer, inventory) into one PostgreSQL warehouse.
  • Designed a star-schema model with slowly changing dimensions, enabling year-over-year and trend reporting without manual snapshots.
  • Cut dashboard refresh times by ~50% by diagnosing slow queries with EXPLAIN ANALYZE and restructuring joins, filters and indexes.
  • Automated Python/pandas profiling of daily OMS, POS and marketplace feeds to block duplicate orders, empty files and invalid values before load.
  • Partnered with finance, merchandising and marketing to define 20+ KPIs, ending conflicting numbers across teams.
  • Owned OMS-to-marketplace reconciliation; documented fixes became the standard reconciliation process.
  • PostgreSQL
  • SQL
  • CTEs
  • Window functions
  • EXPLAIN ANALYZE
  • Python
  • pandas
  • Tableau
  • Git

Baylor Scott & White

Data Analyst

Oct 2022 – Mar 2023

Irving, TX

Delivered healthcare claims analytics across eligibility, provider, benefits and payment systems for the operations team.

Highlights
  • Tracked claims volume, turnaround time, denial rates and payment discrepancies to flag trends for operations follow-up.
  • Built reusable Python/pandas data-quality scripts catching duplicate claims, missing or invalid fields and mismatched member/provider IDs.
  • Reconciled claims, adjustments and payments to source records, confirming reported totals.
  • Maintained weekly Tableau dashboards of claims KPIs and processing trends for the operations team.
  • Ran UAT on new claims reports and documented data definitions and KPI logic to keep reporting consistent.
  • Python
  • pandas
  • Tableau
  • SQL

Relevance lab

Jr Data Analyst

June 2020 – May 2022

Bengaluru, Karnataka

Supply-chain, sales and inventory analytics, data ingestion and reporting automation for business stakeholders.

Highlights
  • Reduced delivery time by 25% by analyzing shipment data in Python to pinpoint supply-chain delays and recommending process changes.
  • Automated data extraction and recurring reporting pipelines, saving ~9 hours of manual work per week.
  • Built API-based ingestion from the inventory processing application into a Delta Lake on AWS S3, the team's central inventory store.
  • Built Tableau dashboards tracking supply-chain delays by product and daily revenue by region.
  • Led analysis projects from design to production, improving data accuracy and customer satisfaction by ~20%.
  • Python
  • SQL
  • PostgreSQL
  • Delta Lake
  • AWS S3
  • APIs
  • Tableau

Kone

Data Analyst Intern

Aug 2019 – May 2020

Chennai, Tamilnadu

Cleaned operational data and built trend reporting for internal stakeholders.

Highlights
  • Cleaned and standardized operational data in Excel and wrote SQL queries validated against existing reports.
  • Built pivot-table summaries and charts of monthly operational trends and presented recommendations to internal stakeholders.
  • Simplified data collection, reducing rework caused by inconsistent entries.
  • Excel
  • SQL
03

Projects

Swipe through projects →

Schematic: generator feeds a transaction list into price-volume-mix, elasticity and regional discount functions, then a report

Pricing analytics

marginbridge

A price-volume-mix margin bridge that reconciles exactly, plus price elasticity estimated with the discount confounder controlled, scored against synthetic data where the true elasticity is known.

Key results
  • Bridge on 6,231 transaction lines reconciles exactly: volume +341,470, mix −275,020, price −37,935.
  • Mix shift explains 88% of the margin headwind and price erosion 12%.
  • Corrected elasticity intervals cover the true value in every segment; naive estimates sit near zero.
  • Python
  • OLS regression
  • Confidence intervals
  • unittest
  • GitHub Actions
Source
Schematic: generator and study split feed uplift learners and cross-fitted nuisance models into policy and evaluation

Customer analytics

upliftpolicy

Retention targeting by uplift instead of churn propensity, with Qini curves and doubly-robust policy evaluation that surfaces customers contact drives away.

Key results
  • Top 30% by uplift adds 4.47 points of retention; top 30% by churn score loses 0.23 points.
  • Sleeping dogs are 37% of the churn-ranked list vs 4% of the uplift-ranked list.
  • T-learner uplift correlates +0.638 with the true treatment effect; 34 tests.
  • Python
  • Logistic regression
  • Causal inference
  • JavaScript
  • Node.js
  • GitHub Pages
SourceLive demo
Schematic: seasonal simulator feeds a classifier and a damped controller, capped by a hard spend ceiling

Marketing analytics

pacingloop

Real-time campaign pacing with hour-of-week median/MAD baselines and a damped, rate-limited controller that avoids false alarms on normal seasonal peaks.

Key results
  • 4/4 planted incidents detected and classified with 0.00 false alarms per week.
  • Damped controller: 279 adjustments and 33 reversals vs 488 and 117 for a naive proportional controller.
  • MAD baseline moves only from 965 to 970 when incidents enter history, vs standard deviation 770 to 3465.
  • TypeScript
  • Node.js
  • Robust statistics
  • Anomaly detection
  • GitHub Pages
SourceLive demo
Schematic: fixtures and candidate links load a consent-gated graph in SQLite, queried by purpose and evaluated for precision

Customer data

consentgraph

Identity resolution where per-purpose consent gates the graph join itself, with match quality measured as precision against planted ground truth and a household-confusion metric.

Key results
  • 1,432 source records resolved against 472 real people across 300 synthetic households.
  • Marketing purposes hold 1.0000 precision with zero household confusions.
  • Service purpose trades precision (0.8458) for 1.000 recall, reported rather than averaged.
  • TypeScript
  • Node.js
  • SQLite
  • Recursive CTEs
  • GitHub Pages
SourceLive demo
Schematic: a column catalogue drives lineage discovery, an erasure plan, a certificate and an adversary that tries to re-identify the subject

Data governance

erasure

GDPR erasure driven by column-level lineage across multiple data stores, verified by an adversary that tries to re-identify the person who was just deleted.

Key results
  • Lineage discovery finds 16 PII columns across 5 stores.
  • Naive table-list erasure leaves the subject identifiable; lineage-driven erasure does not.
  • Dry run by default, producing a certificate of tables touched and rows affected; 17 tests.
  • TypeScript
  • Node.js
  • SQLite
  • Data lineage
  • GitHub Pages
SourceLive demo
Schematic: stage definitions build content-addressed cache keys, run in parallel, and store outputs in a local or S3 cache

Data pipelines

dagcache

A content-addressed, parallel pipeline runner for reproducible workflows that hashes code, params and data, explains why a stage is stale and traces artifact lineage.

Key results
  • SHA-256 cache keys over code, params and input content, with early cutoff to stop needless re-runs.
  • 165 tests with mypy --strict and a coverage gate.
  • About 4,600 lines of typed Python with local or S3 remote caching.
  • Python
  • YAML
  • AWS S3
  • mypy
  • uv
  • GitHub Actions
Source
Schematic of the ratevault bitemporal rate tables and quote engine

Insurance pricing

ratevault

A motor insurance quote engine with bitemporal rate tables, so any past quote reprices exactly using the rates and rules in force at the time.

Key results
  • Backdated correction shows GBP 1995.19 as quoted vs GBP 1671.64 as corrected: GBP 323.55 owed back.
  • Factor-by-factor explanation of every premium; overlap protection enforced by database triggers.
  • Twelve tests run in CI on Node 24.
  • TypeScript
  • Node.js
  • SQLite
  • Bitemporal modeling
  • GitHub Pages
SourceLive demo

More projects

Data integrity

permitflow

A municipal permitting workflow where a declared statechart compiles to database constraints and triggers, so illegal status transitions are rejected by the database itself.

Key results
  • 11 states and 15 declared transitions compiled into a transition table and triggers.
  • Swept all 95 illegal pairs: 95 rejected, 0 accepted.
  • Append-only history; a raw-SQL bulk import that skips the application is rejected too.
  • TypeScript
  • Node.js
  • SQLite
  • PostgreSQL
  • SQL triggers
  • GitHub Pages
SourceLive demo

Offline-first sync

pickpath

An offline-first warehouse picking PWA where reconnect conflicts become supervisor exception tasks instead of silently overwritten picks.

Key results
  • 500 contested SKUs: 500 exceptions raised and 0 silent losses, against 500 silent losses for last-write-wins.
  • Queue replayed twice: 1000 duplicate submissions recognised, outcome unchanged.
  • Each exception task carries the SKU, picker, reason, quantities and the other pickers holding it.
  • TypeScript
  • Reservation ledger
  • Idempotent apply
  • GitHub Pages
SourceLive workbench

Resilience testing

lastknowngood

A backup recovery drill harness that measures real RTO and RPO by running restores and tests backup immutability by attempting deletion in a sealed lab.

Key results
  • Drill 1 measured an RTO of 10h 00m against a claimed 4h 00m; an undocumented credential cycle cost the difference.
  • After breaking the cycle, the same estate met the 4h 00m RTO.
  • A lock configured but not enforced is caught by attempting deletion; achieved RPO of 0h 57m with 12 records lost.
  • Python
  • Drill orchestration
  • Immutability probes
  • GitHub Actions
Source

Retail analytics

Retail Analytics: Customer Segmentation & Revenue Optimization

Analysis of retail transaction data to understand customer purchasing behavior, seasonal sales trends and product category performance, turned into recommendations for revenue, targeting and inventory planning.

Key findings
  • 46–55 age group contributes the highest revenue; Saturdays generate maximum revenue.
  • Q4 recorded the highest sales while Q3 showed a seasonal dip.
  • Revenue differences driven by purchase volume, not pricing; medium baskets give the highest average transaction value.
  • Python
  • pandas
  • NumPy
  • Matplotlib
  • Seaborn
  • Jupyter Notebook
Source

More repositories, including smaller experiments, live on GitHub.

github.com/Nimisharani
04

Skills

SQL & Databases

  • SQL
  • PostgreSQL
  • CTEs
  • Window functions
  • Complex joins
  • Subqueries
  • Query tuning
  • EXPLAIN ANALYZE
  • Execution plans
  • Indexing

Python

  • Python
  • pandas
  • NumPy
  • boto3
  • Data profiling
  • Data cleaning
  • Transformation
  • Workflow automation

Data Modeling & Cloud

  • ETL/ELT
  • Staging and warehouse layers
  • Star schema
  • Fact/dimension design
  • Slowly changing dimensions
  • Source-to-target mapping
  • Delta Lake
  • AWS S3
  • EC2

BI & Visualization

  • Tableau
  • Calculated fields
  • Parameters
  • Interactive dashboards
  • Excel
  • Pivot tables
  • KPI reporting
  • Data storytelling

Statistics & Analytics

  • KPI design
  • Hypothesis testing
  • Confidence intervals
  • Segmentation
  • Correlation
  • Trend analysis
  • Root-cause analysis

Data Quality & Testing

  • Completeness checks
  • Duplicate checks
  • Integrity checks
  • Source reconciliation
  • Business-rule validation
  • UAT test design

Practices

  • Git
  • Technical documentation
  • Reproducible workflows
  • Requirements gathering
  • Stakeholder communication

Domains

  • Retail and e-commerce
  • Marketplace
  • Inventory
  • Returns
  • Healthcare claims
  • Supply chain

Contact

Let's make your numbers trusted.

Looking for an analyst who turns multi-source data into KPIs teams trust? Reach out by email or connect on LinkedIn.