0
0
0

HELLO, I'M

ARYAN PAGARIA

AI / ML ENGINEER

BUILDING
INTELLIGENT SYSTEMS

Exploring generative AI and turning research ideas into practical engineering.

CONTACT ME

02 / ABOUT

BEYOND THE CODE.BUILDING INTELLIGENT SYSTEMS.

I'm Aryan Pagaria. My interests sit at the intersection of AI/ML, research, and the software systems that turn ideas into practical implementations.

I focus on LLM engineering and pipelines, AI agent workflows, and orchestration. I’m interested in understanding how these pieces work together—not just what a model can do on its own.

Aryan Pagaria
ARYAN PAGARIA AI / ML ENGINEERING
EDUCATION
B.Sc. Computer Science with Artificial IntelligenceSRM Arts and Science College
CURRENT
AI Engineer InternSpazor Labs

HOW I THINK

  • 01 / RESEARCHUnderstand before building.
  • 02 / SYSTEMSThink beyond the model.
  • 03 / IMPLEMENTATIONMake ideas work in practice.

03 / SELECTED WORK

ENGINEERING INTELLIGENCE INTO REAL SYSTEMS.

SELECTED PROJECTS · RESEARCH · APPLIED AI

01 / LLM ADAPTATION
QLoRA 4-BIT BASE + LoRA ADAPTERS TRAIN → INFER → EVALUATE

LLM ENGINEERING / FINE-TUNING

Parameter-Efficient LLM Fine-Tuning for Financial QA

Adapted Qwen2.5-3B-Instruct for financial question answering using QLoRA and 4-bit quantization.

  • Built a reproducible FinQA training pipeline with tokenization, gradient accumulation, scheduling and checkpoint recovery.
  • Implemented evaluation using Exact Match, Normalized Exact Match and Numerical Accuracy.
  • Python
  • PyTorch
  • Hugging Face Transformers
  • PEFT
  • Qwen2.5-3B
VIEW ON GITHUB ↗
02 / RETRIEVAL SYSTEMS
RAG DOCUMENT → CHUNKS → EMBEDDINGS FAISS RETRIEVAL → ANSWER

LLM ENGINEERING / RETRIEVAL SYSTEMS

Document Question Answering with RAG

Built a document question-answering system combining retrieval-augmented generation with FAISS similarity search.

  • Developed document ingestion, chunking and embedding workflows.
  • Exposed modular retrieval and question-answering endpoints through FastAPI, with Docker-based execution.
  • Python
  • Hugging Face
  • FAISS
  • FastAPI
  • Docker
VIEW ON GITHUB ↗
03 / SCIENTIFIC COMPUTING
REGRESSION CMS OPEN DATA → JET MASS NLL LOSS + PREDICTIVE UNCERTAINTY

RESEARCH / DEEP LEARNING

Jet Mass Regression with Uncertainty Estimation

Developed deep learning models for jet mass prediction with uncertainty estimation using CMS Open Data.

  • Used negative log-likelihood loss to model predictive uncertainty.
  • Evaluated RMSE, pull distributions and prediction coverage.
  • Python
  • PyTorch
  • CMS Open Data
VIEW ON GITHUB ↗

04 / EXPERIENCE

BUILDING BEYOND THE EXPERIMENT.

PROFESSIONAL EXPERIENCE · AI ENGINEERING

Spazor Labs

JUL 2026- DEC 2026

AI Engineer Intern

Working on LLM-powered systems and agentic workflows, focusing on the engineering behind reliable AI applications.

  • Contributing to LLM pipelines, prompt and context engineering, API/tool integration, and structured outputs.
  • Exploring agent state management, evaluation workflows, and system reliability.

AREAS OF WORK

  • LLMs
  • AI
  • APIs
  • RL

05 / TECH STACK

TOOLS I USE TO BUILD INTELLIGENT SYSTEMS.

LANGUAGES · MODELS · FRAMEWORKS · INFRASTRUCTURE

01 /

PROGRAMMING

The language behind the pipelines.

  • Python
  • Java

02 /

MACHINE LEARNING & DEEP LEARNING

Training, regression and evaluation.

  • PyTorch
  • Deep Learning
  • Model Training
  • Regression
  • Model Evaluation

03 /

LLM ENGINEERING

Adaptation, retrieval and context.

  • Hugging Face Transformers
  • LLMs
  • Reinforcement learning
  • QLoRA
  • Fine-Tuning
  • Retrieval-Augmented Generation (RAG)
  • Prompt and Context Engineering

05 /

BACKEND & API ENGINEERING

Modular interfaces for AI workflows.

  • FastAPI
  • REST APIs
  • Modular Application Architecture

06 /

CLOUD & DEVOPS

Reproducible local execution and version control.

  • Docker
  • Git
  • Kubernetes
  • CI / CD
  • MLOPs

06 / CONTACT

LET’S BUILD SOMETHING INTELLIGENT.

AI / ML · LLM SYSTEMS · ENGINEERING

GitHub LinkedIn Emailaryanpagaria7@gmail.com
VIEW RÉSUMÉ