# Amdu Berga

Amdu Berga is a remote worker on Himalayas, Freelance AI and Data Engineer at AiGot S.r.l., based in Italy.

Source: https://himalayas.app/@amduberga

```text
AI and Data Engineer at AiGot who built Brand Atlas and raised tone classification accuracy from about 28% to about 78%.
```

| Detail | Value |
| --- | --- |
| Availability | Open to opportunities |
| Location | Italy |
| Authorized to work in | Italy |
| Salary expectations | 40k+ EUR |
| Job categories | AI Data Engineer |
| Skills | Large Language Models, RAG, LLM Evaluation, Agentic AI, Machine Learning, Self Supervised Learning, Contrastive Learning, Human Activity Recognition, Time Series Modelling, Prompt Engineering, Fine Tuning, Natural Language Processing, Model Compression, Knowledge Distillation, Adversarial ML |

## What I'm looking for

```text
I’m looking for a research-minded internship where I can work on LLM features, evaluation, and open-model fine-tuning. I’m available for remote work in Europe.
```

## About

```text
At AiGot S.r.l., I built Brand Atlas, a production RAG and agentic LLM platform using ChromaDB, Qwen3 on AWS Bedrock, LangGraph, and FastAPI. I raised tone classification accuracy from about 28% to about 78% by iterating on prompts and measuring each change.

I also added SQLite feedback logging, diagnosed three production bugs, and migrated the LLM backend from Gemini to AWS Bedrock. At EveryWare Lab, University of Milan, I completed a research internship on machine learning for smart-home sensor data.

For my MSc thesis, I tested masked self-supervised transformers for cross-home human activity recognition under 10%, 5%, and 1% label budgets. Transformers reached Macro-F1 0.82 against 0.44 for the BiLSTM baseline in the Milan 10% label setting, and I’m preparing a manuscript on learning-based dynamical systems modelling with transformers.
```

## Experience

### Freelance AI and Data Engineer

AiGot S.r.l.. May 2026 - Jul 2026

```text
Built Brand Atlas, a production RAG and agentic LLM platform, improving tone classification accuracy from about 28% to about 78%. Migrated the LLM backend to AWS Bedrock and handled deployment via EC2 and SSM.
```

Skills: RAG, LangGraph, ChromaDB, Qwen-3, AWS, fastAPI, EC2, SSM, SQLite, XGBoost, SMOTE, Python

### AI Research Intern

EveryWare Lab, University of Milan. Mar 2025 - Aug 2025

```text
Conducted research on machine learning for smart-home sensor data in the lab that later hosted the MSc thesis.
```

Skills: Machine-Learning, Smart-Home-Sensor-Data, Research

### Technical Officer

Commercial Bank of Ethiopia. Dec 2022 - Feb 2024

```text
Worked on ETL pipelines and machine-learning-based credit scoring, ensuring data correctness and traceability. Completed the bank's IS Graduate Trainee Program covering credit analysis, risk, and compliance.
```

Skills: ETL, Machine-Learning, Credit-Scoring, Data-Traceability

## Education

### University of Milano-Bicocca

Master of Science, Artificial Intelligence for Science and Technology. 2023 - 2026. Grade: 101/110

```text
MSc in Artificial Intelligence for Science and Technology, joint programme with University of Milan and University of Pavia. Final grade 101/110, thesis on masked self-supervised transformer pretraining for human activity recognition.
```

### Addis Ababa Science and Technology University

Master of Science, Electrical and Computer Engineering. 2020 - 2022. Grade: 3.92/4.0

```text
MSc in Electrical and Computer Engineering with focus on Control and Instrumentation. GPA 3.92/4.0, thesis on PSO-tuned fractional-order PID controller for mobile robot trajectory tracking.
```

### Addis Ababa Science and Technology University

Bachelor of Science, Electrical Engineering. 2015 - 2021. Grade: 3.74/4.0

```text
BSc in Electrical Engineering with focus on Industrial Control. GPA 3.74/4.0, thesis on control and simulation of a two-wheeled inverted pendulum robot.
```

### DeepLearning.AI

Supervised Machine Learning, Machine Learning

```text
Certification in Supervised Machine Learning.
```

### DeepLearning.AI

Neural Networks and Deep Learning, Deep Learning

```text
Certification in Neural Networks and Deep Learning.
```

### DeepLearning.AI

Calculus for ML and Data Science, Mathematics

```text
Certification in Calculus for Machine Learning and Data Science.
```

### University of Alberta

Software Architecture, Software Engineering

```text
Certification in Software Architecture.
```

## Tech stack

- SQLite
- React
- Python
- MATLAB
- TensorFlow
- PyTorch
- scikit-learn
- FastAPI
- Transformers
- Linux
- Git
- Docker
- SQL
- Google Cloud Run
- XGBoost
- Hugging Face
- LangChain
- ChromaDB
- Pydantic
- Faiss
- LangGraph
- Prompt Engineering

To contact Amdu, message them from this profile with a Himalayas employer account: [Hire on Himalayas](https://himalayas.app/hire).

## People also viewed

- [Vishu S](https://himalayas.app/@vishus): Research Fellow, Università degli Studi di Napoli Parthenope
- [Krys Dessolis](https://himalayas.app/@krysdessolis): Data Scientist, MC Engineering
- [Irene De Nevi](https://himalayas.app/@irenedenevi): Big Data Engineer, Capgemini
- [Elvedin Dobrnjic](https://himalayas.app/@elvedindobrnjic): Software Development & AI Integration Intern, Ikon Srl
- [Juan Sánchez](https://himalayas.app/@juansnchez): Platform Development and Process Automation, Mecanografica
- [Eugenio Carpentieri](https://himalayas.app/@eugeniocarpentieri): Machine Learning Engineer, Capgemini
- [Karim Ellithy](https://himalayas.app/@karimellithy): AI Intern, Instant Software Solutions
- [Fabio Puddu](https://himalayas.app/@fabiopuddu): Senior Data Scientist / AI Engineer, BIP Group
- [Nikita Olefir](https://himalayas.app/@nikitaolefir): Senior Machine Learning Engineer, Avito
- [Matteo Zagano](https://himalayas.app/@matteozagano): RAG Travel Assistant, Independent Project
