Kathy Chhean

🔬 1. Brand Strategy & Positioning (For FunnelUP)

• Market Analysis: Targeted audit of sales funnel experts on Instagram to map the competition, refine the agency's ICP (Ideal Customer Profile), and highlight its uniqueness.
• Visual Identity & Web: Creation of Figma mockups for the agency's own website and design of visual assets for their acquisition channels (Instagram, Social Ads).

🎨 2. Conversion-Focused UX/UI Design (For the agency's clients)

• Sales Funnel Design: High-fidelity design of landing pages and capture pages on Figma for various profiles (sports/mental coaches, technical trainers).
• Growth Synergy: Alignment of UI journeys with the overall content strategy (copywriting, ghostwriting newsletters) to maximize conversion rates (CRO).

🚀 3. No-Code Integration & Mobile-First Approach

• Systeme.io Integration: Rigorous translation of Figma mockups onto the platform, overcoming its technical limitations to preserve aesthetics and usability.
• Responsive Optimization: Complete mobile-first alignment, essential for converting traffic from organic (Instagram) and paid (Ads) acquisition.

🤖 1. Benchmark & Deep Learning

• Architecture Evaluation: Comparative study and implementation of state-of-the-art models: DETR (Transformers via Hugging Face), YOLOv8 (Ultralytics), and RetinaNet.
• Performance Analysis: Evaluation of bounding box accuracy and model robustness against high object density on shelves.
• Technology Choice: Selection of YOLOv8 for its inference speed and relevance in real-time industrial conditions.

📊 2. Low-Data Strategy & Pipelines

• Data Volume Impact: Creation of scripts to test model performance on very limited datasets (from 5 to 40 images per class on the RPC Dataset).
• Data Augmentation: Implementation of image augmentation pipelines (rotations, normalizations) to maximize learning in low-data contexts.
• Metrics Tracking: Rigorous analysis of results using Precision-Recall curves, F1-Score, and mAP (mean Average Precision).

🔮 3. Architecture Recommendations & R&D Vision

• Hybrid Model (Ensemble): Proposal of an architecture separating region-of-interest detection (YOLOv8) and fine-grained product classification (ResNet/EfficientNet).
• Synthetic Generation: Feasibility study to enrich training with virtual data without additional collection costs.

📈 Project 1: Refactoring & UI/UX Optimization of a Dashboard (Python Dash)

• UI/UX Ergonomics: Complete graphical redesign (dynamic callbacks, readability of volumes in billions, harmonization of styles).
• API Anonymization: Integration of multi-selection filters to automate the anonymization of broker flows via the production API (Virtu Financial).
• Data Productivity: Custom module for data extraction and automatic generation of graphical reports (Plotly/PNG) to Excel in one click.
• Data Quality: Normalization and cleaning of historical databases (handling missing values, anomaly treatment).

📊 Project 2: Automation & Decision Support (Bloomberg BQuant)

• Widget Product Ownership: Simplification of the user journey (replacing manual input with interactive dynamic controls and buttons).
• Notification Pipeline: Automatic generation of reports (HTML/PDF) enriched with trader comments for automated email distribution.
• Technical Robustness: Management of complex exceptions (empty dataframes, corporate proxy restrictions, API token authentication).