Built the complete backend powering HeySam’s AI Sales Engineer, seamlessly integrating with platforms like Zoom, Slack, and Google Sheets. Key achievements:

* Led the AI and backend development in close collaboration with the CEO, defining technical direction and architecture for the company’s core product.

* Built scalable APIs and ingestion pipelines to handle real-time Zoom call streams, Slack contexts, and document corpus for LLM fine‑tuning

* Developed custom data pipelines and prompt-engineering frameworks to fine‑tune LLMs for video analysis and live Q&A, delivering instant, context-aware responses during calls

* Engineered low-latency inference infrastructure with optimized batching, caching, model routing, and monitoring to ensure sub-second response times in live settings.

* Enabled frictionless deployment: turn-key Slack app and Zoom integration deployable within two weeks to support live technical Q&A and call intelligence

* Designed and executed AI transformation roadmaps for 4+ early-stage startups, cutting operational costs and accelerating digital adoption.

* Built custom GPT-based workflow automations, eliminating 15–20 hours/week of manual work across marketing, support, and content creation teams.

* Led strategy for computer vision solutions at a retail analytics startup, improving inventory accuracy by 25%.

* Developed an AI readiness assessment framework to prioritize initiatives with clear, measurable ROI and low implementation cost.

As founding ML Engineer at Gorgias, I built and deployed production ML systems that power the platform’s global customer support, scaling to 100M+ daily queries at sub-100ms latency.

* Collaborated with CEO, CTO, and Product VPs to define and execute AI automation strategy across 10 k + e-commerce clients.

* Architected and launched multilingual NLP models (sentiment, intent, summarization, NER, translation) powering Gorgias’s platform, processing 100M+ queries/day at <100ms latency.

* Enabled $4M+ ARR growth and 15%+ CSAT increase by automating key support workflows and accelerating agent responses.

* Owned full ML lifecycle: model development, scalable MLOps, deployment, and 24/7 monitoring (99.99% uptime).

Contributed to Renault’s datalab, leading end-to-end ML solutions across diverse business needs:

* Supply Chain Optimization: Developed models using graph theory and random forests, reducing vehicle delivery times by 12% and cutting logistics costs by €1M+ annually.

* eHandover (Augmented Reality): Built AR part recognition using YOLOv3 and Blender-simulated data, achieving 95%+ part detection accuracy and deployed to 500+ service technicians.

* Customer Review Analytics: Designed NLP pipelines (topic modeling, Doc2Vec, HDBSCAN) to automatically identify and classify vehicle issues, increasing actionable feedback extraction by 3x.

* Contextual Music Recommendation: Created a deep learning recommendation engine (CNN + RNN) leveraging driver mood and landscape from dual cameras, resulting in 20% higher in-car entertainment engagement.

Developed and deployed end-to-end data pipelines for public transport management software used by major cities (Paris, Rochester, Lima, etc.), enabling advanced analytics and ML:

* Data Integration: Automated extraction, normalization, and consolidation of transit data from 10+ disparate city databases, improving data reliability and accessibility.

* Passenger Load Prediction: Built XGBoost models to forecast real-time vehicle occupancy at each stop, increasing load prediction accuracy by 18% and supporting better fleet management.

* Alighting Prediction: Developed ML models to predict boarding and alighting counts, enhancing route optimization and scheduling, leading to 10% reduction in overcrowding events.