Aryan Chaubey

Data Engineering & Analytics

Rising senior at Texas A&M University studying Data Engineering, with minors in Computer Science and Statistics. I build ML prediction systems, data harmonization pipelines, and the full-stack tooling around them.

──01.experience12ms

01 — Experience

Where I've worked

Building analytics infrastructure and predictive models in industry.

Data Analyst Intern

LTIMindtree·Retail & Consumer Packaged Goods

Present
raw retail data→ETL + modeling→dashboards + forecasts

Building analytics pipelines and predictive models for retail and CPG clients.

pipeline latency−32%
forecast accuracy+18%
  • —Developed dashboards and reporting workflows to surface key business metrics across retail channels.
  • —Optimized ETL pipelines for faster data ingestion and reduced processing latency.
  • —Built predictive models to forecast demand and support merchandising decisions.
──02.stack12ms

02 — Stack

Tools & technologies

The stack behind the pipelines.

$ ls -la ~/stack/

languages/

  • -rw-r--Python
  • -rw-r--TypeScript
  • -rw-r--SQL
  • -r--r--R

ml/

  • -rw-r--XGBoost
  • -rw-r--LightGBM
  • -rw-r--scikit-learn
  • -rw-r--Pandas

data/

  • -rw-r--PostgreSQL
  • -rw-r--PostGIS
  • -rw-r--Redis
  • -rw-r--SQLite
  • -r--r--Parquet

backend/

  • -rw-r--FastAPI
  • -rw-r--Pydantic
  • -rw-r--WebSocket
  • -rw-r--Docker

frontend/

  • -rw-r--Next.js
  • -rw-r--React
  • -rw-r--Tailwind

ops/

  • -rw-r--Prometheus
  • -rw-r--Grafana
  • -rw-r--GitHub Actions
  • -r--r--Tableau
──03.projects12ms

03 — Projects

Selected work

Supply chain twins, prediction engines, and data pipelines.

viewport.active
01

AEGIS

Supply chain digital twin where an LLM orchestrates incidents and every number is computed by deterministic code.

nodes40+
tools8
evals10
LLM AgentsPostGISOptimization
viewport.active
02

NNW Media Intelligence Platform

A private production platform I solely designed, engineered, and operate for creator discovery, live highlight detection, searchable VODs, and media publishing.

creators1M+
followers6K+
likes300K+
Media IntelligenceData EngineeringAutomation
03

ProEdge Analytics

Adaptive over/under prediction engine for NFL, NBA, and MLB with calibrated probabilities and drift-aware retraining.

features200+
sports3
models2
Predictive ModelingMLOpsFastAPI
04

ROSETTA

Retail POS harmonization engine where the LLM infers a mapping once per file format and deterministic transforms apply it to every row.

cost$1.12
tests16
calls/row0
Data EngineeringLLMETL
05

KalshiEdge

Real-time edge signal engine for Kalshi prediction markets, streaming live order book data over WebSocket.

workers4
features15+
sizingKelly
Prediction MarketsMLReal-time
06

PRISM

Local-first NBA prop scanner built on LightGBM quantile models with a FastAPI and React interface.

storageSQLite
modelsquantile
runtimelocal
Sports AnalyticsLightGBMFull-stack
──04.about12ms

04 — About

Beyond the pipeline

The details that don't show up in a résumé.

caliber

Aryan

AC-2026

Beyond data pipelines and models, I'm drawn to the technical details that most people overlook — whether it's shaving milliseconds off hardware latency, tuning an 8000Hz polling rate keyboard, or manually calibrating the displays on a multi-complication mechanical timepiece. The craft is in the margins.