# Yaçine Seybou Siddo
> AI Systems Engineer & Independent Consultant
> I build AI systems that keep working after the demo.
> Portfolio: https://yacineseybousiddo.me
> GitHub: https://github.com/Yacine-ai-tech
> Contact: contact@yacineseybousiddo.me

## Profile
AI systems engineer and independent consultant based in West Africa, operating remotely worldwide. Specializing in production deployment of high-fidelity RAG pipelines, governed agent infrastructure (MCP), and resilient architectures engineered for resource-constrained environments.

## Published systems
- IntelAI | Les tableaux de bord BI classiques noient les décideurs sous des métriques brutes sans isolation des rôles et souffrent d'hallucinations lors des synthèses textuelles. | https://github.com/Yacine-ai-tech/IntelAI
- DocIntel | Les flux documentaires du Sahel (reçus froissés, tampons officiels superposés, contrastes faibles) échouent sur les OCR traditionnels et génèrent des coûts API prohibitifs. | https://github.com/Yacine-ai-tech/DocIntel
- RAGeval | Détection impossible des hallucinations silencieuses et de la dérive des modèles en production sans annotation humaine constante et coûteuse. | https://github.com/Yacine-ai-tech/RAGeval
- AgentKit | Risques de sécurité critiques et évasions de privilèges lorsque des agents autonomes manipulent directement des bases de données de production. | https://github.com/Yacine-ai-tech/AgentKit
- StreamPulse | Le volume de leads non qualifié noie les équipes commerciales et les systèmes de CRM coûtent des milliers de dollars par mois pour des fonctionnalités de classification basiques. | https://github.com/Yacine-ai-tech/StreamPulse
- VoiceFlow | Les interfaces textuelles et les assistants vocaux dominants nécessitent une alphabétisation et des smartphones de dernière génération, excluant des milliards d'utilisateurs. | https://github.com/Yacine-ai-tech/VoiceFlow

## Published claims
- {"EN":"Forecasting backtest","FR":"Forecasting backtest"}: MAE 12.48% (median 9.90%)  | dataset= | N= | verified=true
- {"EN":"Production RAG accuracy","FR":"Production RAG accuracy"}: 71.4% ground-truth accuracy, 0.572 average groundedness  | dataset= | N= | verified=true
- {"EN":"A/B test","FR":"A/B test"}: Confirmed persona-based RBAC actually filters retrieval, not just display  | dataset= | N= | verified=true
- {"EN":"SROIE public benchmark","FR":"SROIE public benchmark"}: 95.0% (57/60) zero-shot  | dataset= | N= | verified=true
- {"EN":"Route A accuracy","FR":"Route A accuracy"}: 92.5-100%  | dataset= | N= | verified=true
- {"EN":"Route B accuracy","FR":"Route B accuracy"}: 77.0% CORD, 100% French/FCFA sample  | dataset= | N= | verified=true
- {"EN":"Cost","FR":"Cost"}: $0.0007-0.0021/doc (Route B) vs $0.0048-0.0122/doc (Route A)  | dataset= | N= | verified=true
- {"EN":"GraphRAG-lite entity coverage","FR":"GraphRAG-lite entity coverage"}: 95.0% (7,488/7,878 records)  | dataset= | N= | verified=true
- {"EN":"Throughput/reliability","FR":"Throughput/reliability"}: 550/550 documents processed successfully (100%) at ~1.1 docs/second  | dataset= | N= | verified=true
- {"EN":"HaluEval-QA validation","FR":"HaluEval-QA validation"}: Consensus accuracy 0.785 [0.725, 0.840], F1 0.786 [0.717, 0.843], ROC-AUC 0.870 [0.818, 0.915]  | dataset= | N= | verified=true
- {"EN":"Panel disagreement signal","FR":"Panel disagreement signal"}: Stdev 0.272 on wrong predictions vs 0.069 on correct ones  | dataset= | N= | verified=true
- {"EN":"Adversarial guardrail cases","FR":"Adversarial guardrail cases"}: 14/14 enforced correctly, reproducible deterministically with zero network or LLM calls  | dataset= | N= | verified=true
- {"EN":"MCP tool-invocation benchmark","FR":"MCP tool-invocation benchmark"}: Average execution 1.8s, P95 3.2s, 20/20 success across all four protocol stages  | dataset= | N= | verified=true
- {"EN":"Classification cascade accuracy","FR":"Classification cascade accuracy"}: 8.3% (keyword only) → 64.6% (+embedding) → 91.7% (full cascade), 0.793 macro-F1  | dataset= | N= | verified=true
- {"EN":"Webhook HMAC verification","FR":"Webhook HMAC verification"}: 100% correct accept/reject (90/90 valid processed, 10/10 invalid rejected)  | dataset= | N= | verified=true
- {"EN":"WER/CER","FR":"WER/CER"}: 2.2% / 0.8%  | dataset= | N= | verified=true
- {"EN":"Action-item extraction","FR":"Action-item extraction"}: P=0.502, R=0.518, F1=0.506  | dataset= | N= | verified=true

## Published deployments
- {"EN":"HyperTech Connect (IoT Mesh)","FR":"HyperTech Connect (IoT Mesh)"} | {"EN":"IoT Network & Edge Gateway","FR":"Réseau IoT & Passerelle Edge"} | {"EN":"Validated capacity of 1,016 devices under WireGuard / MQTT mesh network","FR":"Capacité validée de 1 016 dispositifs sous réseau maillé WireGuard / MQTT"} | {"EN":"Production / Validated","FR":"En production / Validé"}
- {"EN":"HyperFlow (Smart Irrigation)","FR":"HyperFlow (Irrigation Intelligente)"} | {"EN":"Digital Agriculture & IoT","FR":"Agriculture Numérique & IoT"} | {"EN":"Autonomous control loop (Sense → Plan → Act → Verify) deployed on Sahelian agricultural operations","FR":"Boucle de contrôle autonome (Sense → Plan → Act → Verify) déployée sur exploitations agricoles sahéliennes"} | {"EN":"Field Deployed","FR":"Déployé sur le terrain"}
- {"EN":"HyperTech Electronics (Retail & AI Platform)","FR":"HyperTech Electronics (Plateforme Retail & IA)"} | {"EN":"Hybrid Commerce & Operations","FR":"Commerce & Opérations Hybrides"} | {"EN":"~127,000 LOC, 150 tables, 5 microservices, 554 backend tests","FR":"~127 000 lignes de code, 150 tables, 5 microservices, 554 tests backend"} | {"EN":"Operational in Production","FR":"Opérationnel en production"}
- {"EN":"Addax SME (Energy Analytics)","FR":"Addax SME (Analytique Énergétique)"} | {"EN":"Industrial Energy & AI","FR":"Énergie Industrielle & IA"} | {"EN":"Consumption forecasting and industrial energy cost optimization","FR":"Prévision de consommation et optimisation des coûts d'énergie industrielle"} | {"EN":"Delivered & Validated","FR":"Livré & Validé"}

## Published credentials
- IBM AI Engineering Professional Certificate (V3) | IBM | 
- IBM RAG and Agentic AI Professional Certificate | IBM | 
- IBM Full Stack Software Developer Professional Certificate (V5) | IBM | 
- Mathematics for Machine Learning and Data Science | DeepLearning.AI | 
- IBM Data Science Professional Certificate | IBM | 
- IBM Data Science Professional Certificate (V3) | IBM | 
- Google IT Support Professional Certificate | Google | 
- Agile Project Management | Google | 
- Leading Transformations: Manage Change | Macquarie University | 
- AI Product Management Specialization | Duke University | 
- Innovation & Entrepreneurship: From Basics to Open Innovation | 28DIGITAL | 
- Understanding Research Methods |  | 
- Principles of UI/UX Design |  | 
- Introduction to Data Engineering |  | 
- Artificial Intelligence in Marketing |  | 
- Ethical Hacking |  | 
- Associate Data Scientist | Qwasar Silicon Valley | 
- Introduction to Software Engineering | Qwasar Silicon Valley | 
- Introduction to Data | Qwasar Silicon Valley | 

## Published profile metrics
- Exactitude Zero-Shot SROIE: 95.0% | Mesurée sur le benchmark public SROIE (57/60 reçus annotés) via DocIntel
- Garde-Fous Déterministes: 14/14 | 14/14 tests adversariaux validés sans appel réseau ni dépendance LLM via AgentKit
- Moteurs Open-Source Livrés: 6 | Publiés en 2026 avec benchmarks mesurés et limites défavorables documentées
- Mois d'Ingénierie en Production: 18+ | Systèmes déployés pour l'agriculture, l'IoT, l'énergie et l'analytique au Sahel

## Data provenance
> This document is generated from published PostgreSQL records.