📍 Berlin, Germany 🎓 MSc Data Science — FU Berlin 🏆 Core Value Member 2024

Building Intelligent
AI Systems

I am a Full-Stack Data Scientist with 4+ years of experience in FinTech and AI consultancy, specialising in NLP, RAG pipelines, fraud detection, multi-agent systems, and MLOps. I build production-grade ML systems that drive real business impact.

Get In Touch → LinkedIn Kaggle GitHub

✉️

Email

afiacuet19@gmail.com

📍

Location

Berlin, Germany

💼

Current Role

Data Science Consultant

🎓

Education

MSc Data Science, FU Berlin

4+

Years of Experience

2+

Industries Served

6+

ML Systems Shipped

Career Bio

My journey in data science and AI

Professional Summary

Full-stack data scientist with 4+ years of FinTech and consultancy experience, specialising in NLP, RAG pipelines, fraud detection, transaction monitoring, OCR, speech-to-text, time-series forecasting, and MLOps.

Skilled in building production-grade ML models and cloud-based AI solutions that drive data-driven business impact. Successfully delivered POCs, RAG chatbots, and agentic systems, contributing to client wins. Awarded Core Value Member 2024 for contributions in applied AI .

✓
Measurable reduction in financial risk via real-time fraud detection pipelines
✓
Significant reduction in manual effort via cloud-native multi-agent document system
✓
Substantial improvement in KYC compliance via OCR-powered document processing

Current Research

Recently completed my Master's thesis supervised by Dr. Thomas Hoppe (Fraunhofer FOKUS), researched the integration of structured knowledge bases into LLaMA's transformer attention mechanisms using a KBLAM-inspired architecture.

The work focuses on injecting structured knowledge entries directly into attention layers, enabling improved reasoning and open-ended QA without the latency overhead of traditional retrieval pipelines.

PyTorchTransformersKnowledge GraphsHPC / Distributed TrainingLLaMA

Technical Skills

Technologies and tools I work with

🐍 Languages

PythonSQLC++

🤖 LLMs & Agentic AI

LangChainLangGraphLlamaIndexFAISSAzure OpenAIDocling

🧠 Deep Learning

PyTorchTensorFlowKerasHuggingFaceCUDA

⚙️ MLOps & Infrastructure

MLflowDockerFastAPIAzureDatabricksCI/CD

📊 Data & Analytics

PySparkPandasNumPyScikit-learnOpenCVNLTK

🎯 Specialisations

NLPRAG PipelinesFraud DetectionTime-SeriesOCRSpeech-to-TextGraph RAG

Experience

My professional journey

Data Scientist (Working Student)

Publicis Sapient

📍 Berlin, Germany

Nov 2023 – Present

Designed and deployed a cloud-native multi-agent system using LLMs and LangGraph, significantly reducing manual documentation effort for a major client.
Built RAG chatbots integrating Azure Cognitive Search and vector retrieval for enterprise conversational AI solutions.
Designed and optimised Graph RAG pipelines leveraging knowledge graph embeddings for multi-hop reasoning.
Implemented dynamic pricing and churn prediction models for an EnergyTech client with Databricks and PySpark, enhancing forecasting accuracy and strategic decisions.
Collaborated with MLOps engineers to build CI/CD pipelines, ensuring model scalability, observability, and compliance.

LangChainLangGraphLlamaIndexAzure OpenAIFAISSFastAPIMLflowDockerPySpark

Data Scientist

SSL Wireless

📍 Dhaka, Bangladesh

Jan 2023 – Sep 2023

Led end-to-end ML projects for FinTech products including fraud detection, transaction monitoring, and OCR-based KYC.
Delivered a Bengali speech-to-text system for customer service apps, meaningfully improving user engagement and accessibility.
Architected OCR-powered document processing pipelines, substantially improving KYC compliance and reconciliation accuracy.
Built real-time analytics dashboards for transaction monitoring, anomaly detection, and predictive insights.

PythonPyTorchTensorFlowAzureFastAPIDockerSQL

Junior Data Scientist

SSL Wireless

📍 Dhaka, Bangladesh

Aug 2021 – Dec 2022

Deployed real-time fraud detection models for high-volume transaction monitoring, driving a measurable reduction in financial risk.
Built time-series forecasting models to optimise payments and inventory management, consistently reducing stockout incidents across production environments.
Applied geospatial analytics to identify high-value merchants and optimise acquisition strategies.
Automated operational workflows using CNN-based face recognition models, improving security and efficiency.

PythonTensorFlowSQLOpenCVPandas

Projects

Research and personal projects

🧠

KBLAM Knowledge Integration (Thesis)

A study of real-world structural knowledge base integration methods in KBLAM and their impact on LLM response performance. Fine-tuned a LLaMA-based model injecting structured knowledge directly into transformer attention layers.

🗣️

Bengali ASR — BUET CSE Fest 2022

State-of-the-art Bengali speech recognition trained on 400 hours of Common Voice audio. Ranked 6th out of 59 teams at the BUET CSE Fest 2022 Deep Learning Sprint.

💊

Stroke Prediction Web App

Web application designed to predict the likelihood of stroke based on patient health indicators, with model explainability via Dalex.

😷

Covid-19 Vaccine Tweet Sentiment

Combined CNN and BERT models to classify vaccine-related tweets as Positive, Negative, or Neutral using the Covid-19 Vaccine Tweets with Sentiment Annotation dataset.

📄

Bengali Punctuation Correction

Fine-tuned a T5 transformer for Bengali punctuation restoration using a Seq2Seq architecture. Achieves robust restoration performance on Bengali text corpora.

🐱

Image Classification (Cats vs Dogs)

Image classifier built from scratch in PyTorch to classify cats and dogs. Includes full data preprocessing pipeline, custom training loop, and evaluation metrics.

Recommendations

What colleagues and managers say

👤

André Koriath

Former Head of Data Strategy, Science & Engineering

Publicis Sapient

★★★★★

I can highly recommend her for any AI‑focused role. I had the pleasure of working with Afia, and she consistently impressed all of us in the Sapient Data & AI team with her strong analytical mindset, deep technical competence, and remarkable independence. During her time with us, she contributed meaningfully to several applied AI initiatives: both client‑facing, billable work and internal business‑development efforts.

Afia combines both breadth and depth in Data Science, Applied AI, and ML research. She works with minimal input, asks the right questions, challenges assumptions thoughtfully, and conducts additional research to elevate the quality of her solutions.

A great example was a business‑development AI Agent demo she built to showcase GDPR compliance verification across contract documents. She not only delivered the prototype and continuously improved it, but also proactively added a UI from a lawyer's perspective — which significantly strengthened our pitches and sales conversations.

Beyond her technical strengths, Afia is a genuinely great person to work with: open, collaborative, and highly appreciated by everyone on the team. An absolute team player.

👤

Samuel Clampitt

Senior Manager Data Science

Publicis Sapient

★★★★★

We've been fortunate to have Afia as a working student in the data domain at Publicis Sapient. She has contributed meaningfully to a number of initiatives in the applied AI space, including working directly with a client on a billable project while writing her MSc thesis.

In terms of technical competencies, Afia has both depth and breadth; in addition, she's dedicated, gets along with everyone, and demonstrates good judgment. I would confidently recommend her for applied AI, ML research, or adjacent roles.

👤

Maximilian Meingast

Senior Manager Data Science

Publicis Sapient

★★★★★

I had the pleasure of working with Afia, who impressed with her strong analytical mindset and hands-on expertise in Data Science and Analytics. Already during her studies, she gained relevant project experience in both client-facing and internal settings, where she successfully applied her technical skills to deliver meaningful results.

Afia is reliable, proactive, and highly appreciated within the team — not only for her professionalism, but also for her open and transparent way of collaborating with others.

Education

Academic background and achievements

🎓

MSc in Data Science

Freie Universität Berlin

Oct 2023 – March 2026 · Berlin, Germany

Thesis at Fraunhofer FOKUS: Integrating structured knowledge bases into LLaMA transformer attention mechanisms (KBLAM-inspired architecture) to improve open-ended question answering and reasoning without retrieval overhead.

PyTorchTransformersKnowledge GraphsHPC

🏆

Core Value Member Award 2024

Publicis Sapient · Berlin

2024

Awarded for outstanding contributions in applied AI innovation and delivery excellence across client projects.

🥈

Kaggle Competition Rankings

Kaggle · 2022

Competition Highlights