Objective: To foster communication among like-minded users through dynamic
group formation based on their responses to specific questions.
Approach: Employed vectorization and K-Nearest Neighbors (KNN) for initial grouping,
with a recommendation system for refined group alignment.

Objective: To advise individuals on financial actions that enhance their chances of
achieving specific goals.
Context: For an Asian bank aiming to offer actionable advice for long-term financial
planning, such as securing a home purchase in 20 years.
Approach: Predictive machine learning models were developed to suggest optimal
actions (example: obtaining a salary increase of at least 10%).

Objective: To surpass the performance of Amazon Personalize’s Hierarchical Recurrent
Neural Network-based recommendation system.
Approach: Adopted a recommender model based on Factorization Machines, significantly
improving recommendations through targeted feature engineering.

Objective: To predict the likelihood of loan repayment by bank customers, aiding
in their segmentation for tailored communication strategies (email, SMS, or phone
calls).
Achievement: Created a behavioral scoring machine learning model, now incorporated
into Receeve’s collection approach.

Objective: To forecast investors’ forthcoming decisions on company investments.
Approach: Transformed the challenge into a graph-based problem, treating investors
and companies as nodes. Developed a Graph Neural Network to predict potential
investor-company connections, integrating this model into StockFink’s prediction
suite.

Objective: To predict vehicle transportation prices, providing dispatchers with a
reliable basis for pricing.
Achievements: Enhanced the machine learning model’s accuracy by approximately
40% through advanced feature engineering and implementing a two-tiered regression
strategy, combining residual and standard regression techniques powered by XGBoost.

Objective: To detect unusual behavior in a sensor-monitored network.
Method: Established a baseline machine learning model predicting expected sensor
readings, using deviations from this model to flag anomalies. Enhanced detection
accuracy with a Graph Neural Network, analyzing the connectivity between sensors
and observable areas.

Objective: Fraud detection.
Approach: I designed and developed a fraud detection model for a Middle-Eastern
Buy-Now Pay-Later platform. A critical step for this client involved data preprocessing,
during which I employed a graph-based approach to identify cliques of
fraudsters. This was the first successful machine learning project for this client.

Objective: given a spectral image that depicts the value of the sensors’s readings,
classify the signals and identify novelties (anomalies).
Approach: Given that there was not sufficient labeled data, I had to rely on selfsupervised
machine learning paradigms. To satisfy the customer’s request for fast
processing, I utilized one of the YOLO architectures. The system was developed
with the mmyolo framework.

Objective: an LLM chatbot to help with HR-related inquiries.
Approach: I led the development of an advanced Retrieval-Augmented Generation
(RAG) system aimed at improving HR data retrieval processes. This system utilizes
the Milvus Vector Similarity Search Database and OpenAI’s API to efficiently source
and integrate extensive HR-related data, enabling it to respond to a broad spectrum
of HR inquiries. We have also attempted to fine-tune an LLM for this task, and the
technologies used were PEFT (LoRA), and sophisticated Prompt Engineering.