The Foundation

Where Curiosity Met Code

It started with a simple question: can machines think like us? A Computer Science degree gave me the tools. A Master's in Cognitive Systems — at the intersection of AI, neuroscience, and human-computer interaction — gave me the obsession. Not just building software, but building systems that perceive, learn, reason, and adapt.

ResearchFraunhofer IPA

Teaching Robots to See & Learn

At Fraunhofer IPA, I worked on making robots collaborate safely with humans in dynamic environments. I applied contrastive unsupervised learning to extract representations from high-dimensional image data — no labels needed. These latent features supercharged deep reinforcement learning policies, cutting latency and accelerating robot learning. This was AI beyond simulations — real-time, physical, unforgiving.

Computer VisionReinforcement LearningRepresentation LearningRobotics
IndustryNokia

Predicting Hardware Failures Before They Happen

At Nokia, the challenge shifted to predicting chip failures under extreme voltage and temperature conditions — catching problems before final testing. I unified data from multiple sensor sources into a single pipeline, trained models from Random Forest to neural networks, and built dashboards that turned complex QA data into decisions engineers could act on.

Predictive ModelingXGBoostTime SeriesSensor DataTableau
Production MLReply (Volkswagen)

Owning the Full ML Lifecycle

Working with Volkswagen through Reply, I took end-to-end ownership — designing and deploying a complete ML pipeline for automotive sales prediction on AWS. I built ETL pipelines that improved prediction accuracy by ~30%, introduced CI/CD for deep learning experiments with Terraform (cutting iteration time in half), and led a cross-account data migration with zero downtime. Over 1.5 years, I owned requirements, timelines, and execution independently.

AWSETL/ELTCI/CDTerraformEnd-to-End ML
Space SystemsSolenix

Anomaly Detection Where Failure Isn't an Option

Satellite telemetry data. Time-series patterns. Subtle deviations that could signal mission-critical failures. At Solenix, precision wasn't a goal — it was the baseline. I built anomaly detection models that identified system deviations early, contributing to systems where early detection is the difference between mission success and catastrophic loss.

Anomaly DetectionTime SeriesSatellite Telemetry
CurrentSoyoma

Bridging AI and Business Impact

Now as an AI Integration Consultant, I help organizations move from "AI sounds interesting" to "AI is driving results." I design and implement solutions that bridge business needs with ML capabilities — practical, scalable adoption over experimental hype. From RAG pipelines to risk analytics, the focus is always on measurable impact.

AI ConsultingRAG PipelinesLLMsScalable Solutions

Research Depth. Production Rigor.

I move seamlessly between research and production — translating advanced ML concepts into robust, real-world applications that create measurable impact. From robots learning in dynamic human environments to deploying scalable pipelines in production, every role sharpened a different edge of my toolkit.

5+
Years of ML Experience
4
Industry Domains
E2E
Pipeline Ownership
R→P
Research to Production
View Experience Get in Touch