neural_core_v4.5.py ● RUNNING

tamaghna@nordstrom:~/ml-systems$ python neural_init.py --mode=production

[INFO] Initializing ML Engineer v4.5...

✓ PyTorch core loaded (CUDA enabled)

✓ Transformer stack initialized

✓ MLOps pipeline ready (Airflow + K8s)

✓ LLM/RAG framework active

[METRICS] 4+ years experience | 98% deployment rate | 10M+ events/day

[STATUS] Available for opportunities...

█

TAMAGHNA NAG

[ ML ENGINEER × LLM ARCHITECT × PRODUCTION AI ]

Building production ML systems at the intersection of computer vision, LLMs, and scalable MLOps. From research to deployment—architecting AI that ships.

Years Exp

% Deploy Rate

10M+

Events/Day

INITIATE CONTACT → DOWNLOAD RESUME ↓

SCROLL TO EXPLORE MY JOURNEY

WHO.I.AM

My Story

Years Production ML

Deployment Success

10M+

Events/Day Processed

MY JOURNEY IN ML

I'm Tamaghna Nag, a Machine Learning Engineer currently architecting production AI systems at Nordstrom. My journey started with a curiosity about how machines could learn, and it's evolved into building end-to-end ML platforms that process millions of events daily.

From Oxford Vision & Sensor Technology where I built transformer-based CV/NLP models for automotive OEMs, to Turing where I consulted on LLM engineering for Meta and Microsoft, to now leading fraud detection ML at Nordstrom—I've learned that the real challenge isn't just building models, it's making them work reliably in production.

I hold an MSc in Computer Science from Warwick where I researched AI-driven traffic optimization, and a BTech from IEM Kolkata with an 8.86 GPA. But my real education has been in the trenches—deploying PyTorch models to Kubernetes, debugging Airflow DAGs at 2 AM, and learning that 98% deployment success comes from the details.

My mission: Build AI systems that don't just work in notebooks, but transform how businesses operate in the real world.

MY.PATH

Professional Journey

CURRENT

Nordstrom

SDE III (Machine Learning)

Jan 2026 – Present

▸ Architected production fraud/risk ML platform with hybrid scoring (policy + ML) processing 10M+ events/day

▸ Built Airflow DAG pipelines for event ingestion, feature generation, and scoring with BigQuery + S3 backend

▸ Implemented MLOps workflows: versioned artifacts, environment-aware configs, rollback controls, observability

Python Airflow BigQuery Kubernetes MLflow

Turing (Meta, Microsoft)

AI/ML Consultant

Aug 2025 – Present

▸ Delivered LLM engineering workflows for Meta-scale research pipelines and MSFT evaluation systems

▸ Built RAG systems with LangChain, FAISS/Chroma, custom chunking, guarded generation for enterprise use

LLMs RAG LangChain FAISS

Oxford Vision & Sensor Technology

ML Engineer

Oct 2022 – Aug 2025

▸ Built transformer-based CV/NLP models (PyTorch, HuggingFace) deployed via MLflow + K8s for OEMs

▸ Designed agentic AI workflows combining Spark streaming + vision models, reducing manual QA by 50%

PyTorch Transformers MLflow Spark

Capgemini

Data Analyst / SWE

Aug 2021 – Dec 2022

▸ Developed DL anomaly detection for IoT sensor streams, reducing downtime by 30%

PyTorch Docker Kubernetes

WHAT.I.BUILD

Selected Work

PRODUCTION

Fraud Detection Platform

End-to-end ML platform at Nordstrom with hybrid scoring, Airflow orchestration, production observability

98% Uptime

10M+ Events/day

Airflow BigQuery K8s MLflow

LLM/RAG

Document Intelligence

RAG framework with LangChain + FAISS for enterprise retrieval with guardrails

LangChain FAISS HF

GitHub →

COMPUTER VISION

VIN OCR Pipeline

YOLOv8 + OCR hybrid for robust vehicle ID extraction

98% Accuracy

YOLOv8 OCR PyTorch

TRANSFORMER

Tiny-LLM

Custom small-scale LLM from scratch

PyTorch Transformers

GitHub →

GET.IN.TOUCH

Let's Connect

I'm always interested in new opportunities to build production ML systems that solve real problems. Whether you're hiring, collaborating, or just want to talk about LLMs and MLOps—let's connect.

tamaghnanag04@gmail.com

Available for new opportunities

neural_connect.py

$ initiate_contact()

> Establishing connection...

✓ Email channel open