StoryImpactExperienceWorkResearchStackContactGet in touch
Open to select AI engineering & research work

I turn frontier AI research into production systems.

I’m Samarth Rai — a Senior AI, Data & Analytics Engineer. I build enterprise-grade AI that actually ships: LLM automation, deep-learning pipelines and $15M+ data platforms — and I publish the research behind it.

See the workGet in touch
4+ yrs
shipping production ML
3
IEEE publications
Dubai
United Arab Emirates
The story

Most AI never leaves the notebook. My whole job is getting it out.

I started where a lot of engineers do — a Computer Engineering degree, a Data Science minor, and a genuine obsession with the papers coming out every week. The difference is I never wanted the research to stay academic. I wanted it running in front of real users, moving real numbers.

So that’s what I’ve spent the last four years doing. At Chalhoub’s Level Shoes, I architect the data platform behind $15M+ of annual pipelines and build the LLM and generative-AI tools that automate how a luxury retailer buys, merchandises and plans inventory — work that used to eat entire analyst teams.

On the other side of the same brain, I publish. Vision Transformers hitting 99.92% on wildfire detection, few-shot evaluation of vision-language models, measuring how well GPT-4 actually reasons — three IEEE papers and counting. Staying at the frontier isn’t a hobby; it’s how I know what’s ready to ship next.

Research tells me what’s possible. Engineering makes it reliable. I live in the overlap.

Foundation

American University of Sharjah (AUS)

B.Sc. Computer Engineering · Minor in Data Science

Sept 2019 — Dec 2023 · GPA 3.34 / 4.0

Founder & President — Open Source ClubResident Assistant — Student Life Department
By the numbers

Outcomes, not activity.

A few of the numbers that describe what shipping AI at enterprise scale actually looks like.

$0M+

Annual data pipelines architected

GA4 · Oracle EBS · Oracle RMS · retail systems

0.00%

Accuracy on the FLAME2 wildfire dataset

Vision Transformer, near-SOTA

0%+

BigQuery cost reduction

Partitioning, clustering & materialized views

0

IEEE peer-reviewed publications

AAIML · Access · ICALT

Experience

Where I've shipped.

From real-time data services to enterprise AI automation — each step closer to the frontier, in production.

Applied AI at retail scale

Senior Analyst, Data & Analytics

Chalhoub Group — Level Shoes

Jul 2025 — PresentDubai, UAE

LLMsGenerative AIPyTorchdbtBigQueryLookerMMM
  • Architect and run $15M+ in annual data pipelines across GA4, Oracle EBS, Oracle RMS and proprietary retail systems with dbt and Python.
  • Built LLM & Generative-AI automation for buying, merchandising and inventory planning — replacing manual analyst workflows and reclaiming significant operational capacity.
  • Applied deep-learning models for market intelligence, surfacing demand patterns and competitive signals from large-scale unstructured data.
  • Shipped Marketing Mix Modeling (MMM) pipelines quantifying channel-level ROI that directly shape growth-marketing budget allocation.
  • Cut BigQuery costs 25%+ by redesigning GA4 event tracking with optimized partitioning, clustering and materialized views.
  • Delivered Looker & Looker Studio dashboards adopted as the single source of truth across Finance, Marketing, CRM and Supply Chain.
Backend & platform

Software Engineer

Chalhoub Group

Mar 2024 — Jul 2025Dubai, UAE

Django RESTPostgreSQLDockerAWS EKSOracle EBS
  • Engineered scalable microservices with Django REST Framework and PostgreSQL for internal financial data automation.
  • Built intercompany transfer-pricing automation integrated with Oracle EBS — saving 50 man-days every quarter.
  • Containerized services with Docker and deployed on AWS EKS, improving deployment speed by 60%.
Real-time data & AI features

Software Engineer

Beno Technologies

May 2023 — Mar 2024Dubai, UAE

FastAPIAWS LambdaPythonAI Chatbot
  • Built analytics APIs and real-time data-ingestion services with Python FastAPI and AWS Lambda.
  • Shipped an AI chatbot and data-visualization features integrated with live microservices — lifting user engagement 25%.
Selected work

Systems people actually use.

Enterprise AI automation and peer-reviewed research — a cross-section of what I've built and shipped.

Chalhoub · Level Shoes2025

LLM & GenAI Retail Automation

A suite of LLM-powered agents that automate the retail lifecycle — buying, merchandising and inventory planning — turning manual analyst workflows into reviewed, auditable automations.

Analyst hours reclaimedBuying · Merch · InventoryHuman-in-the-loop
#llms#generativeai#agents#rag
Research project2024

Wildfire Detection — Vision Transformer

A Vision Transformer trained to detect wildfires from aerial imagery, reaching 99.92% accuracy on the FLAME2 benchmark — pushing near state-of-the-art on a safety-critical task.

99.92% accuracyFLAME2 datasetViT architecture
#visiontransformer#pytorch#computervision
Lufthansa2024

Aircraft Fault-Detection MLOps

An end-to-end MLOps pipeline for aircraft fault detection — Go for the serving layer, PyTorch for the models and Flutter for the on-the-ground inspection app.

GoLang servingPyTorch modelsFlutter client
#mlops#golang#pytorch#flutter
Chalhoub · Level Shoes2025

Marketing Mix Modeling Pipelines

Statistical + ML pipelines that quantify channel-level ROI across the marketing stack, giving growth teams a defensible basis for budget allocation.

Channel-level ROIStatistical + MLBudget allocation
#mmm#python#statistics#ml
IEEE AAIML 20262026

Crack Detection — Incremental Learning

A deep-learning crack-detection system deployed as an MLOps pipeline that keeps learning incrementally from new data without full retraining — published at IEEE AAIML.

Incremental learningProduction MLOpsPeer-reviewed
#deeplearning#mlops#incrementallearning
Chalhoub · Level Shoes2025

GA4 → BigQuery Cost Redesign

Re-architected GA4 event collection and BigQuery storage — eliminating redundant events without losing granularity and cutting warehouse spend by more than a quarter.

25%+ cost cutNo loss of granularityPartition + cluster
#ga4#bigquery#dbt#costeng
Research

Publishing at the frontier.

Peer-reviewed work across deep learning, vision-language models and LLM evaluation — the reading list that keeps my engineering honest.

01
IEEE AAIML2026

Deep Learning Based Crack Detection MLOps Pipeline using Incremental Learning

doi:10.1109/AAIML67890.2026.11498221

Read paper
02
IEEE Access2025

Few-Shot Evaluation of Vision Language Models for Detecting Visual Defects in Autonomous Vehicle Software Requirement Specifications

IEEE indexed
03
IEEE ICALT2024

Measuring Fluency, Coherency and Logicality of GPT-4 Generated EGRA Comprehension Stories

IEEE indexed
Toolkit

The stack I reach for.

From the model layer down to the infrastructure it runs on — end to end, so nothing gets lost between the paper and production.

01

AI / Machine Learning

  • LLMs
  • Generative AI
  • PyTorch
  • Hugging Face
  • Vision Transformers
  • Deep Learning
  • MLOps
  • RAG
  • Agents
02

Data & Analytics

  • Python
  • SQL
  • dbt
  • BigQuery
  • Looker
  • Looker Studio
  • GA4
  • MMM
03

Infrastructure

  • AWS
  • Airflow
  • Docker
  • Kubernetes
  • Git
  • FastAPI
  • Django REST
Let’s build

Have an AI problem worth shipping?

Whether it’s LLM automation, a deep-learning pipeline or a data platform that needs to scale — I’d love to hear about it. Based in Dubai, UAE, working with teams anywhere.

rai@samarthrai.comLinkedIn

+971 56 695 2628 · Dubai, UAE