Erwan Folquet

AI-Native Leadership & R&D Recovery

• Operate as a one-person product team, orchestrating multi-agent AI development workflows (Claude Code, Codex) to

design, build, and ship end-to-end across backend, data, and UI/UX, delivering the scope of full engineering and design

teams without the headcount.

• Advise company owners on practical AI adoption, including which tools and workflows to use and how to apply them,

turning the AI shift from a threat into a competitive edge.

• Serve as the senior technical authority for founder-led clients, turning ambiguous business goals into concrete

architecture, scope, and technical decisions.

• Led the full SR&ED claim with KPMG (refundable federal and Quebec R&D credits), building append-only

documentation that made R&D verifiable under audit and recoverable as cash.

Generative Media Platform

• Built a production generative-image platform on a provider-agnostic abstraction over Google Gemini and Higgsfield, compositing a real product into AI-

generated scenes by resolving vision context (product facts, palettes, brand directives).

• Designed a brand-learning pack compiler: a pipeline that ingests unstructured brand assets and emits structured, versioned creative rules making on-brand output a deterministic, reusable asset.

• Business impact: collapsed creative time-to-market from days to minutes and self-serve, cutting content cost by

replacing external production with in-house on-brand generation.

Real-Time Conversational Voice System

• Built an AI voice-screening platform that filters candidate pools without manual first-round calls: recruiters share one

link, candidates complete a spoken screen on their own time, and recruiters get structured reports to decide who to call.

• Engineered human-feeling conversation on LiveKit and Gemini Realtime (WebRTC, sub-second

latency), with questions generated from each job description, seniority-calibrated across 50+ roles and conducted within

the recruiter's and company's guidelines.

• Implemented and optimized AWS Bedrock for enterprise-scale generative AI applications, integrating knowledge bases

with OpenSearch Serverless to enable advanced RAG (Retrieval-Augmented Generation) capabilities for customer-

facing LLM applications (OpenAI, Anthropic Claude, DeepSeek).

• Architected the new Discount Pricing system responsible for 80% of company revenue. Optimized pricing strategies

during sales periods, boosted revenue with increased gross margin, minimized overstock, improved execution time by

30%, significantly reduced cloud costs, drove scalability, profitability, and solidified the company’s competitive edge.

• Key architect of the AI/ML Hub system, reducing AI/ML models prototyping, development, and deployment time from 4

months to 2 weeks. Streamlined workflows, accelerated innovation, and improved delivery efficiency (SageMaker,

Lambda, Step Function, Cloudformation, IAC, CICD, Bash, Jenkins, Docker, S3, Athena, ECR, ECS, Event Bridge, API

Gateway, OpenAPI, Route 53, VPC).

• Designed and deployed a scalable Search Engine Management system, optimizing Google Ads campaign expenditures

and enabling the Marketing team to manage a multi-million-dollar annual budget effectively, leading to measurable cost

savings and increased ROI.

• Built a data-driven Inventory Buying Optimization system, reducing overbought inventory by 20%, lowering operational

costs, and providing actionable insights for more efficient decision-making to the Buyers team.

• Built an enhanced Observability and Analytics framework for AI/ML workflows by integrating SageMaker, CloudWatch

custom metrics, Slack alert notifications and Datadog, enabling real-time monitoring and faster debugging cycles.

• Delivered scalable production-grade AI/ML solutions with architectures supporting real time (live), serverless endpoints,

or batch inference, ensuring seamless integration with existing business systems.

• Optimized the Content Recommendation system serving millions of users decreasing churn rate, improving conversion from guest users, and increasing users spendings (Spark, Spark ALS, Databricks, Cloudformation, Step Function).

• Migrated the Recommendation system data processing pipeline from legacy to microservices reducing overload issues, legacy platform dependency, and data processing time (SQL, Snowflake, Databricks, AWS).

• Designed a Content Moderation system to automatically tag and filter NSFW content in user uploaded images and videos (Computer Vision, Keras, Resnet, OpenCV, Pillow, Cuda).

• Created a Credit Card Fraud detection system reducing chargebacks by 30% (SciKit Learn, XGBoost).

• Streamlined data pipelines for faster AI ML model development.

• Data mining, cleaning, and preprocessing of structured and unstructured data (images and videos).

• Wrote documentation of scripts, AI ML models and guides for cross team usage.

• Maintained communication flow with stakeholders both non technical (business/high level), and technical (developers).