Daniel Amirifard | AI/ML Engineer

About

I am an AI/ML Engineer with 4+ years of experience in Generative AI, Large Language Models, and ML/DL domains. I graduated from Politecnico di Milano and have delivered 30+ AI projects across academic and industry environments.

My work focuses on building scalable AI systems, LLM applications, RAG pipelines, agentic workflows, and production-ready model deployments. I have worked on practical AI solutions connected to Hermes/Baker Hughes, STMicroelectronics, Nokia, and Polimi Data Scientists Association.

I enjoy transforming complex technical ideas into reliable products — from RAG-based invoice classification and multi-step LLM agents to neural architecture search, computer vision, and forecasting systems that improve automation, speed, and decision-making.

Experience

Nov 2025 — Present

Artificial Intelligence Engineer · Hermes / Baker Hughes

Full-time · Milan, Italy

Developed 3+ production-oriented GenAI solutions from PoC to deployment, using RAG pipelines, LangGraph, and Amazon Bedrock to improve automation, scalability, and business impact.

Built agentic AI systems with multi-step reasoning, workflow automation, prompt engineering, and decision-making pipelines, reducing manual analysis by around 85% across client-facing business processes.

Delivered AI applications across financial invoice classification, chemical-sector workflows, and energy-sector use cases, with one ML classification pipeline achieving a 93% F1-score.

RAG
LangGraph
Amazon Bedrock
Mistral
Agentic AI

May 2025 — Nov 2025

Machine Learning Engineer · STMicroelectronics

Full-time · Milan, Italy

Designed and deployed automated neural architecture search workflows for LLM agent bots, improving model accuracy and computational efficiency by 40%.

Implemented zero-cost proxy optimization methods, accelerating generative AI benchmarking by 10× while lowering resource consumption.

Constructed and integrated an AutoNN LLM agent that automatically designs neural network structures, reducing model architecture generation time by 70%.

NAS
LLM Agents
AutoNN
Zero-Cost Proxies
GenAI

Feb 2023 — Oct 2024

Machine Learning Engineer · Polimi Data Scientists Association

Part-time · Milan, Italy

Developed, trained, and optimized 10+ machine learning models across computer vision, anomaly detection, and time-series forecasting.

Organized data pipelines for efficient ETL processes and sourced relevant datasets adopted in 5+ projects including SIIT and breast cancer detection.

Showcased project progress to team members and international students biweekly, contributing to the completion of four significant GenAI projects.

Computer Vision
Anomaly Detection
Time Series
ETL
GenAI

Sep 2021 — Sep 2022

Data Scientist · Doran Smart-City

Full-time · Tehran, Iran

Executed machine learning algorithms including Random Forest and XGBoost to analyze and model urban traffic data, resulting in a 15% improvement in traffic flow.

Redesigned predictive models using gradient boosting and random forests for smart energy management, leading to a 10% reduction in energy consumption within city infrastructure.

Random Forest
XGBoost
Traffic Analytics
Smart Energy

Education

Sep 2022 — Mar 2026

MSc, Telecommunication Engineering

Computer Science and Communications · Politecnico di Milano

Milan, Italy

Sep 2016 — Sep 2021

BSc, Engineering

Amirkabir University of Technology

Tehran, Iran

Projects

Featured Work

Interactive AI project showcase

Each project is organized around its goal, measurable achievement, and tools used — highlighting RAG, agentic AI, computer vision, AutoML, SaaS architecture, and production deployment impact.

01 / 08

FIN

RAG · Invoice Classification

RAG-Based Financial Invoice Classification Pipeline

Goal

Automate milestone classification from invoice text fields using a production-oriented RAG and agentic workflow.

Achievement

Built an AI-powered invoice classification pipeline using RAG, FAISS, LangGraph, and Mistral via Amazon Bedrock, achieving a 93% F1-score.

Tools Used

Amazon BedrockRAGFAISSLangGraphMistral

02 / 08

RAG

NLP · Question Answering

Long Form Question Answering ChatBot

Goal

Build a long-form question answering system able to retrieve relevant context and generate structured answers using transformer models.

Achievement

Created the NLP pipeline in a team of four using BART, LLAMA, and RAG, securing a ROUGE score of 22.

Tools Used

RAGLLAMALoRABARTHugging Face

03 / 08

NAS

AutoML · Zero-Cost Proxy

Zero-Cost Proxies in AutoNN LLM Agent

Goal

Estimate neural architecture performance without expensive full training, supporting faster architecture search inside an AutoNN LLM agent.

Achievement

Integrated zero-cost proxies and achieved 0.78 Kendall’s Tau on CIFAR-x datasets for reliable training-free architecture ranking.

Tools Used

Google CloudAzure MLGenAIFine TuningOpenAI

04 / 08

BOT

LLM · Pharmaceutical AI

SIIT ChatBot

Goal

Develop a chatbot to support pharmaceutical synthesis workflows and help users access technical information more efficiently.

Achievement

Spearheaded a technical team of 10 with JEMP startup for SIIT Company, reducing production time by 25% through chatbot-assisted workflows.

Tools Used

RAGLangChainGPTBERTFAISS

05 / 08

ART

Computer Vision · Face Parsing · SaaS

AI-Powered String Art Portrait Generator

Goal

Convert user-uploaded portrait images into clean, printable DIY string-art instructions through an end-to-end AI and web application pipeline.

Achievement

Architected an AI SaaS pipeline that transforms portraits into up to 8,000 optimized pin-to-pin string-art instructions using face parsing, edge detection, grayscale enhancement, circular board mapping, and greedy optimization.

Tools Used

Computer VisionFace ParsingFastAPIDockerSaaS

06 / 08

3D

Computer Vision · Autonomous Driving

Cooperative Perception in Automated Cars

Goal

Build a V2V cooperative perception system for automated driving by fusing LiDAR-based 3D object detection features across connected vehicles.

Achievement

Constructed a cooperative perception pipeline using OPV2V, LiDAR, PointPillars, ROI filtering, and attention fusion, achieving 0.914 AP@0.5 and 0.837 AP@0.7 for 3D object detection.

Tools Used

OPV2VLiDARPointPillarsROI FilteringTransformers

07 / 08

TS

Forecasting · Deep Learning

Traffic Forecasting

Goal

Forecast traffic and channel behavior using deep learning models for time-series prediction and network performance estimation.

Achievement

Directed a team of three and reached 9% MAPE in estimating 50 Mbits/s channels using the GEANT dataset with RNN and Conv-LSTM networks.

Tools Used

Time SeriesRNNLSTMConv NetsMLflow

08 / 08

AI

Data Science · Vision · Security

Other AI & Data Science Projects

Goal

Explore applied machine learning problems across classification, clustering, computer vision, intrusion detection, and forecasting.

Achievement

Delivered projects including ClimateFish Clustering, Mac-Derandomization, Application Identifier, Breast Cancer Detection, ML Intrusion Detection, and Plant Classification.

Tools Used

R-CNNYOLOGANsVAEsMLflow

Auto sliding every 30s

Skills

Programming

PyPython

SQLSQL

GitGit

GHGitHub

Machine Learning & AI

PTPyTorch

TFTensorFlow

KKeras

SKScikit-learn

PdPandas

NpNumPy

SpSciPy

MtMatplotlib

Generative AI & LLMs

RAGRAG

AIAgentic AI

LCLangChain

LGLangGraph

PEPrompt Engineering

BOTChatbots

Cloud & DevOps

AWSAWS

BRAmazon Bedrock

S3Amazon S3

AZAzure ML Studio

DKDocker

APIFastAPI

CICI/CD

LXLinux

Workshop Experience

Jun 2024

Impact of Non-Gaussian Noise on LDPC Performance · Nokia

Vimercate, Italy

Presented to over 100 attendees at Nokia, showcasing a simulation of non-Gaussian noise effects on LDPC performance using QAM 4, 16, and 64.

LDPC
QAM
Simulation
Technical Presentation

Mar 2024

MultiNet-Route-Optimizer · Nokia

Vimercate, Italy

Presented a 65% improvement in routing efficiency for a Nokia-defined problem, achieved by a four-person team using top-down and bottom-up optimization strategies.

Route Optimization
Network Systems
Team Project
65% Efficiency Gain

Certifications

Neural Networks and Deep Learning

DeepLearning.AI

Supervised Learning

DeepLearning.AI

Linux LPIC 1

Tehran Institute of Technology

Python

Tehran Institute of Technology

Get In Touch

I am open to AI engineering opportunities, research collaborations, and projects involving LLM applications, RAG pipelines, agentic AI systems, computer vision, AutoML, and scalable deployment.

Send Email Connect on LinkedIn