Available for PhD & Research Collaborations

Papa Ahmadou
Seydou Sow

AI Research Engineer

I build reliable and resource-efficient machine-learning systems for data-scarce environments, with applications in computer vision, climate AI and geospatial intelligence. Based in Dakar, Senegal, I am open to international PhD opportunities and research collaborations.

Papa Ahmadou Seydou Sow
Flood AI
Computer Vision
Remote Sensing
19yrsYears of reanalysis used
94.7%Vision model F1-score
40k+Images in identity AI
4Deep architectures benchmarked
01 · About

Research-driven AI engineer

I am an AI Research Engineer based in Dakar, Senegal, focused on building rigorous, reproducible and explainable machine-learning systems that work under real-world, data-scarce constraints.

Growing up in Senegal, I saw first-hand how AI tools built for ideal laboratory conditions often break under real constraints, limited data, limited compute and limited connectivity. Closing that gap is what motivates my research.

My work spans computer vision, climate AI and geospatial analytics, from flood early-warning systems and railway predictive maintenance to biometric identity pipelines. I care deeply about Human-Centered AI and AI for Sustainable Development.

I hold a double-degree M.Sc. in Artificial Intelligence and Smart Technologies from UTT (France) and Université Iba Der Thiam de Thiès (Senegal), graduated with honours (mention Bien), and I am seeking doctoral and research opportunities.

Curious and driven by challenges, I enjoy exploring innovative approaches, learning new methods and designing AI solutions tailored to concrete problems.

Research Interests

  • Computer Vision & Visual Recognition
  • Climate AI & Environmental Monitoring
  • Geospatial & Remote Sensing
  • Multimodal Machine Learning
  • Human-Centered AI
  • Early Warning Systems
  • AI for Sustainable Development
Research Vision

Building AI that works where it is hardest

My long-term goal is to develop trustworthy, explainable and resource-efficient AI for environments where data, compute and connectivity are scarce, with a particular focus on Africa. I care about systems that are not only accurate, but deployable, auditable and genuinely useful on the ground.

Axis 01

Explainable & reliable ML

Models whose decisions can be inspected, validated and trusted in high-stakes settings such as health, climate and critical infrastructure.

Axis 02

Computer vision under constraints

Robust visual recognition with small, imbalanced and low-quality data on limited hardware, from remote sensing to biometrics and video analysis.

Axis 03

AI for decision support

Recommendation, forecasting and monitoring systems that help real users, citizens, clinicians, scouts, make better data-driven decisions.

Application domains: climate resilience and early-warning systems, health diagnostics, critical infrastructure, and sports analytics & talent detection, the latter explored through my applied platform BibaFoot.

02 · Featured Research

Master Thesis: Saytu-Mbeund

AI for Flood Prediction & Detection in Senegal

Research question: how can machine learning and computer vision support reliable flood early warning under data scarcity and limited computational resources?

Supervisor: Prof. Ibrahima Gueye · École Polytechnique de Thiès (EPT) · Defended 2026 · grade 16/20

  • Forecasting: 1-to-3-day flood-risk prediction with leakage avoidance and forward-chaining temporal validation; the final model reaches AUC-ROC 0.852 with a numerical weather forecast.
  • Vision benchmark: grouped five-fold cross-validation of ResNet-18, EfficientNet-B3, Swin-Tiny and ConvNeXt-Tiny on 386 deduplicated citizen images, with Grad-CAM.
  • Integration: a transparent two-level pre-alert and image-confirmation workflow, reported as a scenario analysis rather than a field-validated system.
Climate AIComputer VisionRemote Sensing Explainable AIGeospatial Analytics
Read full thesis ↗
Forecasting model · 1–3 day horizon
AUC-ROC · reanalysis only0.791
AUC-ROC · + weather forecast0.852
Raw-forecast alert rule0.840
Final forward-chaining evaluation · 2024–2026
Vision model · ConvNeXt-Tiny
F1-score94.7%
AUC-ROC0.981
Flood precision93.7%
Best of four backbones · grouped five-fold CV
AUC-ROC of the flood forecasting model with and without numerical weather forecast
Contribution of the weather forecast: the same 2005–2023 model goes from AUC-ROC 0.791 to 0.852 when the numerical forecast is added (+0.061, 95% CI 0.045–0.078), forward-chaining evaluation 2024–2026.
ROC curves of the four CNN and transformer backbones for flood image detection
Visual flood detection: out-of-fold ROC curves of the four backbones on 386 images (grouped five-fold CV); ConvNeXt-Tiny reaches AUC-ROC 0.981.
03 · Featured Project

ANIP Benin 2025: Identity AI System

AI-powered identity document verification, built for the ANIP Benin Challenge 2025 (Agence Nationale d'Identification des Personnes). Finalist. Final presentation at the Palais des Congrès, Cotonou, November 2025. The system implements three complementary biometric tasks; the value lies in a rigorous, honest methodology under real-world constraints (CPU-only training, small and imbalanced datasets).

Task 1

Face Matching & Verification

Facial recognition and identity verification on official documents. Iterated from a Siamese network (contrastive / BCE) to metric learning with Triplet Loss + hard-negative mining, then to MTCNN face alignment + FaceNet embeddings with 6-view Test-Time Augmentation and FAR-calibrated thresholds (V2 → V10).

Cosine similarityROC threshold calibrationMTCNN alignment
Face verification calibration: ROC, precision-recall and similarity distribution
TensorFlowPyTorchFaceNetOpenCV
Task 2

Age Estimation

Deep-learning regression for age estimation from facial photos, trained on 40,012 images (17,417 identities). EfficientNetB0 backbone with transfer learning, 30 epochs and ReduceLROnPlateau scheduling.

Best val MAE: 9.16 yrs200 test images
Exploratory data analysis of the age dataset
TensorFlowEfficientNetB0Transfer Learning
Task 3

OCR & Document Fraud Detection

Five-class fraud classification on identity documents from 4 countries (arizona_dl, esp, est, rus). Two versions: EfficientNetB0 baseline (v1) → EfficientNetV2B0 + Cosine Decay with advanced augmentation (v2), CPU-only.

5-class · 10,000 train imgsBest val acc: 20.4%
Confusion matrix of the 5-class fraud detection model
TensorFlowEfficientNetV2OCR

Key learning: real-world biometric AI is genuinely hard: face verification needs robust embedding spaces, age estimation degrades on small/imbalanced data, and 5-class fraud detection with limited data resists easy gains. The value is in the methodology, not just the metrics.

04 · Experience & Education

A research trajectory

Dec 2025 – Sep 2026

Computer Vision Research Engineer

Absar Group

AI-based predictive maintenance of railway infrastructure. Investigated single-stage detection (YOLO family) for structural defect identification, tackling class imbalance and small-object detection; built reproducible annotation & dataset-curation protocols and ran ablation studies.

Jan – Jun 2025

AI & MLOps Engineer

Engy Partner

Personalisation & recommendation for educational technology. Designed an ML-based pedagogical recommendation module and built the full experimental + production stack (data pipeline, backend, deployment), bridging research prototypes and Human-Centered AI.

2022 – 2026

M.Sc. Artificial Intelligence & Smart Technologies

UTT (France) · UIDT Thiès (Senegal) · Double Degree

Graduated with honours (mention Bien). M1: 12.69/20 · M2: 15.56/20 · Thesis defence: 16/20.

2019 – 2022

B.Sc. Computer Science, Software Engineering

Université Iba Der Thiam de Thiès, Senegal

Foundations in software engineering, algorithms and systems.

05 · Research Projects

Selected work

Railway Defect Detection

Empirical study of YOLO-based pipelines for automated identification of railway track defects. The pipeline combines YOLOv8 detection, OpenCV segmentation and a rule-based logic engine to flag displaced fasteners under realistic operating conditions.

YOLOv8OpenCVIndustrial AIPredictive Maintenance

ANIP Benin Identity AI Challenge 2025

Finalist. A 3-task biometric system: Siamese face matching, deep age estimation and 5-class document-fraud detection across 4 countries. See the dedicated breakdown below.

BiometricsMulti-Task LearningOCR
See full breakdown ↓

Real-Time Climate Data Platform

Streaming ingestion & processing architecture (Apache Spark Streaming) for high-frequency environmental data as a substrate for climate-aware decision-support systems.

Climate AIStreaming DataDecision Support

Satellite Remote Sensing & Vegetation Indices

Geospatial experimentation from my flood-mapping work in Senegal: I exploited Sentinel-2 spectral bands (B03, B04, B08 and SWIR B11/B12) to compute water and vegetation indices (NDWI, MNDWI, NDVI) and study before / during / after flood periods over Keur Massar, Kaolack and Kaffrine. Built in Python with rasterio and GeoTIFF rasters, plus Google Earth Engine and geemap. The same spectral-index pipeline, NDVI in particular, is directly relevant to vegetation monitoring and can support agricultural mapping when combined with suitable data and methodology.

Explored directions
Remote SensingNDVI / NDWI / MNDWIRaster ProcessingGoogle Earth EngineGeospatial Analysis
Stack
PythonrasterioGeoTIFFgeemapSentinel-2QGIS

Credit Scoring · DataTour 2025

Gradient-boosting credit-scoring models (LightGBM, XGBoost, Random Forest) and a neural-network scorer with advanced feature engineering and cross-validated tuning, developed for the national phase of the pan-African data science competition.

Target variable distribution and class proportions
EDA, class imbalance: only 3.36% of cases are defaults, a strong driver of the modelling strategy.
Top 25 feature importances by gain
Feature importance (gain): credit cost rate, utilisation and credit limit lead the top-25 drivers.
Credit scoring evaluation: ROC curve, confusion matrix and probability distribution
Evaluation: ROC, confusion matrix and predicted-probability distribution on the validation set.
XGBoostLightGBMNeural NetworksFeature Engineering

Mobile Money Fraud Detection · DataTour 2026

Second place at the national ranking (team Hackatsuki) at DataTour 2026: a CatBoost fraud detector for mobile-money transactions, scored on Average Precision. The winning recipe leans on fold-safe target encoding (the sender's historical fraud rate) and strict temporal validation to avoid account-ID leakage, with a CV-to-leaderboard compass to iterate offline and submit only proven gains. Ranked 2nd nationally (Senegal), just behind the champion.

Explored directions
Fraud DetectionGradient BoostingTarget EncodingTemporal Validation
Stack
PythonCatBoostLightGBMXGBoostscikit-learnPandas
View on GitHub ↗

AI-Based Malaria Detection

CNN classifier for malaria parasite detection on blood-smear microscopy images, deployed as a real-time web interface to assess AI-assisted diagnostics in low-resource clinical settings.

Medical ImagingAI for HealthDeep Learning

Carton QC · YOLO11 Defect Detection

A YOLO11 object detector for automated visual quality control of cardboard boxes: it localises and classifies each box as normal, cracked, opened or wet (anything not normal is flagged as a defect), trained on ~3,000 annotated images across four classes. It reaches mAP50 88.4% and mAP50-95 79.3% on the held-out test set, served through a Streamlit app for real-time inspection.

Explored directions
Object DetectionDefect ClassificationIndustrial VisionData Curation
Stack
YOLO11UltralyticsPyTorchOpenCVRoboflowStreamlit

ResearchOS · Multi-Agent Research Assistant

An open-source "operating system" for scientific research: 14 specialised AI agents run as pipelines or drag-and-drop workflows, a multi-model router picks the LLM by cost, speed, quality and privacy (OpenAI, Anthropic, Gemini, Ollama, Groq and more), and a hybrid BM25 + dense RAG answers with citations. Live literature search across arXiv, OpenAlex, Semantic Scholar, CrossRef and HAL. Used in my own research: it verified the DOIs of my flood-forecasting article's bibliography. Flask + Celery/Redis + PostgreSQL back-end, React 19 front-end, 96 automated tests.

Explored directions
LLM AgentsRAGMulti-Model RoutingClean Architecture
Stack
PythonFlaskReact 19PostgreSQLCelery / RedisSocket.IO
ResearchOS literature search: 29 real articles retrieved from CrossRef, HAL and OpenAlex for flood forecasting ResearchOS workflow builder: research, summarizer, citation and reviewer agents chained ResearchOS agents catalogue: 14 specialised agents Literature search across 5 sources · agent workflow builder · 14 specialised agents

Svrai · AI Fact-Checking from Video

Paste a video, audio file, link or text: Svrai transcribes it (Whisper, plus a Wolof speech model for local content), extracts the factual claims with an LLM, and checks each one against official sources through 8 source plugins (World Bank, OpenAlex, CrossRef, Wikidata and more) and targeted search on official domains. Each claim gets a transparent verdict (corroborated / uncorroborated / contradicted / uncertain) with a confidence level, citations and reasoning, never the model's memory alone. FastAPI + React, user accounts, 25 automated tests.

Explored directions
LLM ReasoningRetrieval & SourcingSpeech-to-TextClaim Verification
Stack
PythonFastAPIWhisperLLM (Claude)yt-dlp / ffmpegReact
Svrai verification path: claim, World Bank source, official figure, deterministic calculation and verdict Svrai home screen: text, image or video input Svrai verification report with per-claim verdicts and confidence Traceable verdict: the figure comes from the World Bank and the comparison is computed, not generated by the LLM
Open live ↗

BibaFoot · Football Intelligence Platform

A live football-management platform used as a testbed for applied AI: recommendation, performance analysis and decision support for players, clubs and recruiters.

ReactFastAPIFirebasePostgreSQLDockerMinIO
Open live ↗

Honours, Awards & Competitions

Master with HonoursDouble-degree M.Sc. AI, UTT (France) & UIDT Thiès (Senegal), 2026.
DataTour 2025 · National PhasePan-African data science competition; built gradient-boosting credit-scoring models (LightGBM, XGBoost, RF) with advanced feature engineering during the national phase.
ANIP Benin Identity AI Challenge 2025National AI challenge on civil identity; face-embedding, age-estimation, OCR & fraud-detection pipeline on >40k images.
DataTour 2026 · 2nd Place (Senegal)Pan-African data science competition; 2nd of the national leaderboard (team Hackatsuki) on mobile-money fraud detection with a leakage-safe CatBoost pipeline (Average Precision).
06 · Technical Skills

Tools of the trade

Research Stack

PythonPyTorchTensorFlowScikit-Learn HF TransformersOpenCVYOLOConvNeXt ResNetEfficientNetXGBoostLightGBM spaCyPandasNumPyApache Spark rasterioGeoPandasGoogle Earth EngineQGIS

Reproducibility & Deployment

Git / GitHubDockerFastAPIFlask Hugging Face HubRoboflowAWSGCP PostgreSQLMongoDB

Methods

Transfer LearningTransformersMultimodal Fusion Object DetectionImage ClassificationTime-Series Forecasting Recommender SystemsOCRModel EvaluationAblations Remote SensingSpectral IndicesGeospatial Analysis

Certifications & Languages

Generative AI (Simplon)AI Fundamentals (FORCEN) Data AnalyticsCybersecurityMobile Dev
French
Fluent
English
Professional
Wolof
Native
07 · Contact

Let's build AI for impact.

Open to PhD positions, research collaborations and applied AI engineering roles, especially in computer vision, climate AI and AI for sustainable development.

Dakar, Senegal