Albert Tchaptchet Womga
AI & Machine Learning Specialist | Data Scientist | Expert
⭐ Albert TCHAPTCHET WOMGA
AI & Machine Learning Specialist | Data Scientist | Expert
🚀 About Me
I am a results-driven AI & Machine Learning Specialist with 15+ years of experience in high-dimensional data analysis, predictive modeling, NLP, computer vision, and time-series/sensor fusion.
I build real-world AI solutions across education, healthcare, legal services, IoT systems, and business automation. I recently completed a Post-Graduate Diploma in AI & Machine Learning, strengthening my expertise in deep learning, Spark/Hadoop, transformers, and applied ML engineering.
I’m passionate about:
✔ Turning raw data into intelligent systems
✔ Behaviour detection from multimodal sensors
✔ Real-time analytics & forecasting
✔ Building automation systems
✔ Applying AI in education, healthcare & legal domains
🔥 What I’m Working On (2024–2025 Projects)
🚧 Ongoing Projects (In Progress)
🌾 CSIRO Image2Biomass Prediction (In Progress)
github.com/womgaalbert/csiro-image2biomass-prediction
RGB image-based pasture biomass prediction using deep learning & multimodal fusion.
Challenge: Kaggle CSIRO Image2Biomass Challenge
Target: RMSE < 300 kg DM/ha
Approach:
- CNN ensemble (EfficientNet, ConvNeXt, Swin, ViT)
- PEFT (LoRA) for fine-tuning
- FiLM-based metadata fusion (NDVI, height, site, season)
- Multi-task learning (biomass + NDVI + height prediction)
- Advanced augmentation (CutMix, MixUp, perspective correction)
- 5-fold GroupKFold cross-validation
- TTA + isotonic calibration + LightGBM stacking
Tech Stack:
PyTorch, TIMM, Albumentations, OpenCV, scikit-learn, LightGBM
Status:
Training ensemble models — Expected RMSE ~285 kg DM/ha.
🧩MAP Project Deployment (In Progress)
github.com/womgaalbert/MAP-Charting-Student-Math-Misunderstanding-Deployment
Deploying the MAP misconception detection system as a web-ready service:
- API design using FastAPI
- Web app built with Streamlit
- Docker containerisation for portability
- Real-time student input inference
- Educator-facing UI for misconception insights
💼 Technical Expertise
| Category | Tools & Technologies |
|---|---|
| Languages | Python, SQL |
| ML/DL Frameworks | PyTorch, TensorFlow, Keras, scikit-learn |
| Time-Series & Sensors | LSTM, GRU, Transformers, feature engineering |
| NLP | NLTK, Transformers |
| Computer Vision | OpenCV, CNNs |
| Big Data | Spark, Hadoop, PySpark |
| Data Tools | Pandas, NumPy, SciPy, Tableau |
| Domains | Education AI, Healthcare analytics, Legal automation, IoT analytics |
🧰 My Stack & Tools
|
|
|
|
|
|
🔥 Completed Projects
📘 1. MAP – Charting Student Math Misunderstandings
🔗 GitHub: github.com/womgaalbert/MAP-Charting-Student-Math-Misunderstandings
AI system detecting K–12 math misconceptions from short student explanations.
Highlights:
- Transformer-based text classification
- Error typology & conceptual reasoning patterns
- Feature engineering for linguistic markers
- Evaluation using multi-class F1-score
- Full training + production-ready pipeline
🧠 2. ARC-2025 Hybrid – Abstract Reasoning (CNN + DSL)
🔗 GitHub: github.com/womgaalbert/arc-2025-hybrid
Hybrid AI system designed for ARC reasoning tasks integrating:
- CNN / U-Net for visual grid feature extraction
- Symbolic reasoning engine (DSL)
- Program induction to solve unseen tasks
- Top-3 accuracy optimization
- Data augmentation for grid-based reasoning
✈️ 3. FlightRank 2025 – Aeroclub RecSys Cup
🔗 GitHub: github.com/womgaalbert/FlightRank-2025
Personalised recommendation system for business-traveller flights.
Core Features:
- Hybrid collaborative filtering + feature ranking
- Group-wise ranking model
- Optimised for HitRate@3 metric
- Feature engineering: routes, time windows, traveller intent
- Complete training pipelines, inference, and leaderboard submission
🎧 4. BFRB Gesture Classification – Sensor Fusion
🔗 GitHub: github.com/womgaalbert/BFRB-Detection-Sensor-Fusion
Multimodal time-series classification for detecting Body-Focused Repetitive Behaviours (BFRBs).
Modalities:
- IMU accelerometers
- Thermopile temperature sensors
- Time-of-Flight distance sensors
Techniques: - LSTM / GRU / Transformer models
- Sequence-to-label classification
- Sliding window segmentation
- Sensor-fusion + normalisation pipelines
- Macro F1-score optimisation
🔋 5. Energy-ARIMA-Forecasting
🔗 GitHub: github.com/womgaalbert/Energy-ARIMA-Forecasting
Time-series forecasting of U.S. industrial energy production.
Methods:
- ARIMA / SARIMA
- KPSS / ADF tests
- Box-Cox transformation
- Multi-step forecasting
- RMSE, MAE, MAPE evaluation
Includes full notebook, visualisations & diagnostics.
📸 6. ConvNet CIFAR-10 Classifier
🔗 GitHub: github.com/womgaalbert/ConvNet-CIFAR10
Custom CNN trained from scratch for CIFAR-10 image classification.
Components:
- Convolution layers + Batch Norm
- Dropout regularisation
- Data augmentation
- AdamW + Cosine LR schedule
- Training visualisations + accuracy & loss curves
📸 🔢 7.Handwritten Digit Classification – MLP vs. CNN
Side-by-side comparison of MLP and CNN on MNIST, each tested with Dropout and Gaussian Noise regularization. Components:
- Baseline MLP & CNN architectures
- Dropout (rate=0.5) regularization
- Gaussian Noise injection (stddev=0.2)
- Adam optimizer + categorical crossentropy
- Accuracy/loss curves + confusion matrix + misclassification analysis
🧭 Professional Experience
Data Scientist – CM Avocats (Quebec, Canada)
- AI automation & intelligent scheduling Data-Centric Financial Management and Accounting
- NLP for legal document drafting
- Design and maintain robust legal data backup and security protocols
Data Scientist – Ministry of Secondary Education (Cameroon)
- National-scale education analytics for 5,000+ schools
- EMIS data integration & dashboards
- ML for policy optimization and student performance
- Predictive models for resource allocation
Biostatistician – Faculty of Medicine of the University of Yaounde I (Healthcare)
- Predictive clinical modeling
- AI-supported biostatistics
- Pediatric health analytics
- Research protocol design
🎓 Education & Certifications
🎓 Post-Graduate Diploma in Artificial Intelligence & Machine Learning
CIMT College, Canada
🎓 Master’s Degree in Applied Statistics
National Advanced School of Engineering, Cameroon
🎓 Diploma in Mathematics
University of Yaoundé
🎓 Certifications
- Udacity – Programming for Data Science
- DataQuest – Python for Data Science
📊 GitHub Stats
🤝 Let’s Connect
📧 Email: albtchap@gmail.com
🔗 LinkedIn: https://www.linkedin.com/in/albert-womga-009a7931/
📍 Ottawa, Ontario, Canada
