# Taha Baziz — AI Technical Program & Product Lead // ML Systems Architect

> AI Technical Product & Program Lead with a rigorous CPGE mathematics foundation and proven execution bridging frontier ML research, quantitative optimization, and production infrastructure. Currently at the LVMH Holding AI Factory leading enterprise agentic systems and Stanford HAI research collaborations. Previously Equativ’s first ML Product Manager (high-frequency RTB bid optimization, -60% CO2, +250% DL revenue uplift) and AI Deployment Strategist across 13 CAC40 enterprises. Expert in agentic orchestration (ADK, LangGraph), low-latency inference, model evals, and rapid SOTA prototyping.

- **Location:** Paris, France
- **Email:** taha@tbaziz.com
- **Site:** https://aegis-portfolio-834287843746.europe-west1.run.app
- **Generated:** 2026-09-16

## Experience

### AI Program Manager & Delivery Lead — LVMH

*May 2024 - Present · Paris, France · AGENTIC AI / ENTERPRISE INFRA*

Operating at LVMH Holding inside the AI Factory with direct reporting to C-Level executives (CPO, CDO, CTOs, CSO, CISO). Spearheading fast-paced prototyping and deployment of agentic systems across LVMH Maisons de Luxe, vibe coding with Claude Code & Antigravity while maintaining high-security VPC, IAP, ILB, and Cloud Run network architecture under cyber threat environments.

- **Situation:** LVMH needed fast-velocity AI agent innovation across its luxury Maisons while enforcing strict CISO cybersecurity policies, zero data retention on Enterprise Vertex AI, and high-network isolation (VPC, ILB, Private DNS, CDN, MFA, IAP).
- **Task:** Serve as AI Program Manager & rotating AI Product Owner working in peers with Tech Leads and POs, reporting to C-levels (CPO, CDO, CTOs, CSO, CISO) to assess internal use cases, deliver rapid prototypes, and build scalable agentic platforms.
- **Action:** Implemented agentic workflows using ADK, Agent Engine, LangChain, and LangGraph on GCP Cloud Run & BigQuery. Set up telemetry using PostHog, Agent Traces, and W&B. Built Dataiku ML pipelines for LLM data enrichment, clustering, and operational alerting systems. Vibe-coded rapid prototypes using Claude Code and Antigravity.
- **Result:** Accelerated prototyping velocity across LVMH Maisons with zero security breaches, strict data governance, and strategic top-line and bottom-line financial impact.

- **Metrics:** Direct C-Level Impact Across LVMH Maisons & Stanford HAI

- **Stack:** Agentic Systems (ADK, LangGraph), Claude Code & Antigravity, GCP Cloud Run & VPC, PostHog & W&B Telemetry, Enterprise Vertex AI, Dataiku DSS, Stanford HAI

**Stanford Human-Centered AI (HAI) Collaboration & Fast Prototyping**

Collaborating directly with the Stanford Institute for Human-Centered Artificial Intelligence (HAI) on human-centered AI endeavors ('we enhance, we don't replace'). Battle-testing open-source harnesses and vibe coding prototypes using Claude Code and Antigravity. Architecting agentic systems with ADK, Agent Engine, LangChain, and LangGraph on Google Cloud Run & BigQuery with telemetry loops in PostHog, Agent Traces, and W&B.

Domains: Stanford HAI Research, Human-Centered AI, Agentic Systems (ADK, LangGraph), Claude Code & Antigravity, Enterprise VPC Isolation

### AI Strategy Consultant, Engineer & Product Lead — INDEPENDENT AI ADVISORY & PRACTICE

*February 2024 - May 2024 · Paris, France · AI STRATEGY / AGENTS / VOICE AI*

Delivered 360° end-to-end AI advisory across the Paris tech ecosystem — bridging C-Level strategy (auditing cash flows, revenue streams, and cost structures) with hands-on AI engineering, custom integrations, autonomous AI SDRs, operational voice call agents, platform maintenance, and executive win/loss analytics.

- **Situation:** Startups and tech scale-ups in Paris needed full-spectrum AI guidance, from financial cash-flow & cost-structure auditing down to hands-on AI model deployment and operational agent engineering.
- **Task:** Act as 360° AI Partner — diagnosing business strategic opportunities, architecting custom AI tools/agents, deploying cloud pipelines, and maintaining platform analytics.
- **Action:** Audited client revenue models, benchmarked SOTA LLMs and voice providers, built custom AI SDRs and operational call handling agents, and established automated win/loss tracking dashboards.
- **Result:** Successfully delivered production AI agent systems and data pipelines, driving top-line growth and operational efficiency for client leadership.

- **Metrics:** End-to-End C-Level Advisory to Production Deployment

- **Stack:** C-Level Strategy, AI SDR Agents, Voice Call Automation, Voice AI (SingAndClap), Cloud Architecture, Data Management

**Voice AI Benchmarking & Autonomous Sales/Ops Agents**

Engineered custom AI Sales Development Representative (AI SDR) pipelines and operational voice call systems. Developed proprietary benchmark harnesses (including voice.singandclap.com) to evaluate frontier LLMs, sub-second STT/TTS models, and transformation pipelines for Paris tech ecosystem clients.

Domains: C-Level Strategy Consulting, AI SDR Agents, Operational Voice Systems, Voice AI Benchmarks (SingAndClap), Transformation Pipelines

Reference: https://voice.singandclap.com/

### AI Deployment Strategist & FDE Lead — Wonderful.ai

*Dec 2025 - Feb 2026 · Paris, France · VOICE & AGENTIC DEPLOYMENT / CAC40*

Spearheaded strategic deployment of high-scale agentic capabilities on the Wonderful platform, focusing exclusively on +$500k/year ARR enterprise use cases across 13 CAC40 companies. Engaged directly with CAC40 C-Level executives alongside GTM leads to audit technical feasibility, prioritize high-leverage roadmaps, and co-deploy conversational voice/written, back-office, and batch-processing AI agents alongside Forward Deployed Engineers (FDEs) targeting >80% containment rates.

- **Situation:** CAC40 enterprises required high-reliability autonomous AI agent deployments operating at massive scale (+$500k+ annual platform spend per use case) with guaranteed success containment rates.
- **Task:** Serve as AI Deployment Strategist — pairing with GTM to pitch C-Levels, auditing technical feasibility, and building POCs/Demos/Production agents in peers with Forward Deployed Engineers (FDEs).
- **Action:** Benchmarked STT, low-latency LLM, and TTS providers. Architected conversational voice/written agents, back-office automation agents, and high-throughput batch-processing agents. Optimized workflows to reliably achieve >80% autonomous containment.
- **Result:** Successfully unlocked high-scale agentic use cases across 13 CAC40 clients with >80% containment rates and +$500k ARR platform spend validation.

- **Metrics:** 13 CAC40 Clients, +$500k ARR Use-Cases, >80% Containment Rate

- **Stack:** Wonderful Platform, Forward Deployed Engineering, STT/TTS Latency Evals, Conversational Voice/Text Agents, Back-Office Automation, CAC40 C-Level Engagement

**STT, Low-Latency LLM & TTS Enterprise Provider Benchmarks**

Conducted systematic benchmarking across STT engines, ultra-low latency LLMs, and TTS voice synthesis providers for enterprise-wide deployment. Built custom evaluation suites to measure round-trip voice latency, dialect nuances, and batch-agent execution throughput for CAC40 clients on the Wonderful platform.

Domains: CAC40 C-Level AI Strategy, Conversational Voice & Text Agents, Back-Office & Batch-Processing Agents, STT/TTS Provider Evals, Containment Rate Metrics (>80%)

### Lead AI Product Manager — Oventi Consulting

*April 2025 - November 2025 · Paris, France · RETAIL PRICING / AI LAB / MENTORSHIP*

Lead AI Product Manager at Oventi Consulting. Mentored 2 AI PMs at Oventi and 2 Tech PMs inside Carrefour's Data Factory. At Carrefour, scaled AI retail pricing recommendation coverage from 3% to 60% of in-store inventory, achieved a 70% pricer acceptance rate, and deployed an autonomous AI Pricing Agent. Created Oventi's internal AI Lab from scratch, deploying AI agents and LLM pipelines that won >€200,000 in new consulting contracts in under 3 months.

- **Situation:** Carrefour's Data Factory had pricing algorithms stalled at 3% in-store inventory coverage due to poor feedback, while Oventi needed an internal AI Lab practice to win enterprise AI contracts.
- **Task:** Act as Lead AI PM — lead change management at Carrefour to scale pricing adoption, mentor AI/Tech PMs, and establish Oventi's AI Lab capabilities.
- **Action:** Built feedback loops with retail pricers, scaling inventory coverage from 3% to 60% with a 70% acceptance rate. Deployed an autonomous AI Pricing Agent. Created Oventi AI Lab, building AI agents and LLM chatbots for insurance and enterprise clients.
- **Result:** Achieved 60% autonomous inventory pricing coverage at Carrefour, mentored 4 PMs across Oventi/Carrefour, and generated >€200,000 in new consulting revenue for Oventi in under 3 months.

- **Metrics:** 3% → 60% Inventory Pricing Coverage, +€200k Contracts in 3 Months

- **Stack:** AI Pricing Agents, Change Management, Retail Pricing ML, Oventi AI Lab, LLM Chatbots & ETL, PM Mentorship & Leadership

**Retail Inventory AI Pricing Agents & Oventi AI Lab Creation**

Engineered change management and feedback loops for retail pricing algorithms inside Carrefour's Data Factory, deploying an AI Pricing Agent that validates price recommendations against market constraints and profit margins. Built Oventi's AI Lab, architecting AI agents, LLM chatbots, and ETL pipelines for French insurance firms and enterprise clients.

Domains: Oventi AI Lab Creation, Carrefour Data Factory, Retail Inventory Pricing ML, AI Pricing Agent, PM Mentorship, +€200k Contract Generation

### First AI & Data PM (Thiga AI Lab Founder) — Thiga

*June 2024 - March 2025 · Paris, France · THIGA AI LAB / TIIME MISSION*

Thiga's first-ever experienced AI/Data Product Manager. Co-founded Thiga's Data & AI Tribe and internal AI Lab. Built a 300+ consultant mission RAG agent and a sales upselling agent. Ran live web servers (Replit, Retool, Streamlit, Cloud Run) tutoring non-technical PMs on LLM compute, storage, and RAG. On consulting mission at Tiime (accounting SaaS), led data platform re-architecture migrating from RabbitMQ to Kafka for real-time ML inference, saving €320,000/yr in cloud costs (>€400k legacy spend), and establishing Agile A/B testing frameworks and LLM data validation agents.

- **Situation:** Thiga needed to establish its Data & AI practice and organize 12+ years of unindexed consultant mission data, while client Tiime faced >€400k/yr skyrocketing cloud costs and zero Agile/eval practices in ML teams.
- **Task:** Build Thiga's Data & AI Tribe and AI Lab, and serve as Data & ML PM on consulting mission at Tiime to re-architect their data platform and overhaul product delivery methodologies.
- **Action:** Created RAG agents for 300+ consultants and sales upselling. Tutored PMs on LLM compute & RAG. At Tiime, orchestrated RabbitMQ to Kafka benchmarks, implemented manual A/B testing frameworks, and deployed LLM invoice validation agents.
- **Result:** Saved €320,000/yr in projected cloud infrastructure costs at Tiime, established Agile sprint & A/B eval practices, and built Thiga's core AI Lab capabilities.

- **Metrics:** €320k/yr Cloud Cost Savings, 300+ Consultant RAG System Built

- **Stack:** Thiga AI Lab, RAG Agent Systems, Kafka & RabbitMQ, LLM Data Validation, Replit & Retool Prototyping, GCP Cloud Run, Agile A/B Testing

**Thiga 300+ Consultant Mission RAG Agent & Tiime Real-Time ML Platform**

Engineered RAG-based AI agents indexing 12+ years of mission archives across 300+ Thiga consultants. Built sales proposal generation agents. On consulting mission at Tiime, benchmarked event streaming architectures (RabbitMQ to Kafka) for real-time accounting auto-completion ML inference and deployed LLM document validation pipelines.

Domains: Thiga AI Lab Creation, Consultant Mission RAG Agent, Sales Proposal AI Agent, Tiime Consulting Mission, RabbitMQ to Kafka Migration, LLM Data Validation, PM LLM Compute Tutoring

### Machine Learning Product Manager (First ML PM) — Equativ

*Nov 2022 - Feb 2024 · Paris, France · HIGH FREQUENCY RTB / CIR RESEARCH*

Served as Equativ's first dedicated ML Product Manager at the French leader of AdTech & high-frequency Real-Time Bidding (RTB). Managed cross-functional DS, MLE, DE, Full-Stack, and SRE teams, reporting directly to CPO, CRO, and CTO while collaborating with CFO & FP&A. Aligned DS roadmaps with Crédit Impôt Recherche (CIR) scientific R&D subventions, boosting deep learning revenue uplift by +250% (+40% overall revenue in €), cutting server CO2 by 60%, and slashing infra costs by 40%.

- **Situation:** Equativ needed business-aware ML leadership across high-frequency RTB to optimize real-time revenue mechanics while balancing massive bare-metal infrastructure costs and carbon footprint.
- **Task:** Act as Equativ's first ML Product Manager — bridging DS/MLE engineers with executive business leaders (CPO, CRO, CTO, CFO, Sales), optimizing Bid Floors/Prices, and auditing ML roadmaps for Crédit Impôt Recherche (CIR).
- **Action:** Tracked RPMA (Revenue Per Mille Auctions) and RPMBR (Revenue Per Mille Bid Requests) North Star metrics. Conducted complex A/B tests with infra-cost awareness. Built SHAP explainability POCs, implemented CO2-driven traffic shaping heuristics, and deployed NLP contextual domain augmentation. Managed hybrid on-prem GPU/CPU inference + cloud analytics with Prometheus, Grafana, Snowflake, BigQuery, and PostgreSQL.
- **Result:** Boosted deep learning revenue uplift by +250% (achieving +40% net revenue growth in €), reduced CO2 emissions by 60%, cut infra costs by 40%, and validated CIR government research subventions.

- **Metrics:** +250% DL Revenue Uplift (+40% € Net Revenue), -60% CO2, -40% Infra Cost

- **Stack:** High-Frequency RTB ML, SHAP Explainability, CO2 Traffic Shaping, NLP Contextual Ad Curation, Hybrid Bare-Metal/Cloud, Prometheus & Grafana, Snowflake & BigQuery, CIR R&D Subventions

**Real-Time Bidding Bid Floor Optimization, SHAP Explainability & CIR Validation**

Justified scientific novelty of ML roadmaps for Crédit Impôt Recherche (CIR) R&D subventions. Spearheaded 3 major ML breakthroughs: 1) Built SHAP deep learning explainability POC in-product to aid pricing/throttling teams; 2) Engineered CO2-aware traffic shaping heuristics reducing carbon output by 60%; 3) Applied NLP on publisher domains for semantic contextual ad curation without tracking cookies.

Domains: High-Frequency RTB Algorithms, Bid Price & Bid Floor Math, SHAP Explainability POC, CO2-Aware Traffic Shaping, Domain NLP Contextual Augmentation, CIR R&D Scientific Validation

### Data Scientist (Trade Finance & Cash Management) — BNP Paribas CIB

*March 2021 - June 2022 · Paris, France · TIME SERIES / ETL / NLP RISK*

Data Scientist in the Trade Finance & Cash Management team at BNP Paribas CIB. Built time series forecasting models, ETL/ELT data pipelines, data analytics dashboards, statistical models, and automated NLP document processing pipelines for corporate banking analysts.

- **Situation:** BNP Paribas CIB Trade Finance & Cash Management needed data science, time series forecasting, and document automation to process complex corporate banking operations.
- **Task:** Serve as Data Scientist — developing time series models, ETL/ELT pipelines, executive dashboards, and NLP risk analytics for trade finance streams.
- **Action:** Designed statistical time series models, built automated ETL/ELT data pipelines, created interactive dashboards, and deployed NLP entity recognition for contract processing.
- **Result:** Automated 70% of manual document review steps and provided executive data analytics across trade finance operations.

- **Metrics:** 70% Contract Review Automated & Time Series Pipelines Built

- **Stack:** Time Series Models, ETL / ELT Pipelines, Data Analytics & Dashboards, Statistical Modeling, NLP, Python

### Product Data Analyst & Operations (Acting Product Manager) — HireSweet (YC S20)

*August 2020 - February 2021 · Paris, France · CANDIDATE MATCHING / CIR RESEARCH*

At HireSweet ATS (hiresweet.com), served as Product Data Analyst & Operations, stepping into Product Manager responsibilities alongside the COO. Partnered with ML researchers to digest math-heavy NLP papers and build candidate matching algorithms that surpassed LinkedIn search precision, audited under Crédit Impôt Recherche (CIR).

- **Situation:** HireSweet needed to build market-leading candidate sourcing algorithms to beat LinkedIn Sourcing algorithms for tech recruiters.
- **Task:** Operate as Product Data Analyst & Operations (acting PM alongside COO), digest mathematical NLP papers with ML researchers, and design candidate matching models audited under Crédit Impôt Recherche (CIR).
- **Action:** Translated search & embedding research papers into product features, built candidate relevance scoring models, analyzed user funnel ops, and compiled technical CIR documentation.
- **Result:** Delivered sourcing precision beating LinkedIn's search algorithms, resulting in Y Combinator (S20) selection and CIR tax credit approval.

- **Metrics:** Outperformed LinkedIn Search Precision (CIR Audited)

- **Stack:** Product Data Analytics, Operations & Acting PM, Candidate Matching ML, NLP Search Algorithms, CIR Filing

**Candidate Sourcing Algorithms Outperforming LinkedIn (CIR Verified)**

At HireSweet ATS (hiresweet.com), worked with ML researchers on state-of-the-art candidate sourcing algorithms. Read math-heavy NLP and search papers to engineer candidate matching algorithms that surpassed LinkedIn's search precision, audited under the Crédit Impôt Recherche (CIR) institute.

Domains: Candidate Sourcing Algorithms, Outperforming LinkedIn Search, CIR R&D Technical Auditing

Reference: http://hiresweet.com/

## Case Studies

### AI Agent VPC Security Isolation & SRE Telemetry — LVMH

- **Situation:** Deploying autonomous AI agents across decentralized business units created significant security and token overflow risks.
- **Task:** Architect an enterprise isolation framework with full SRE telemetry and zero data leakage.
- **Action:** Built VPC guardrails, automated token budget limits, and implemented SRE telemetry loops.
- **Result:** 100% secure agent deployment with real-time observability.

- **Metrics:** Enterprise Agent Oversight & Zero Security Breaches

### Sub-Second Voice AI Architecture & Dialect Evals — Wonderful.ai

- **Situation:** Enterprise banking customers required sub-second voice interactions without MS Graph API latency delays.
- **Task:** Engineer a real-time streaming voice architecture.
- **Action:** Created WebSockets streaming pipeline and benchmarked TTS providers.
- **Result:** Sub-second response latency achieved.

- **Metrics:** Sub-Second Latency & 100% Banking Compliance

### Retail Pricing Model Coverage Expansion (+60%) — Carrefour

- **Situation:** Pricing ML model rollout was stuck in validation phases across 5 DS and 3 DA teams.
- **Task:** Unblock deployment pipelines and accelerate store coverage.
- **Action:** Re-engineered feature store pipelines and standardized validation protocols.
- **Result:** Coverage expanded by +60% in 1 quarter.

- **Metrics:** +60% Pricing Model Coverage in 1 Quarter

### Cloud Datalake Migration & 90% Cost Cut — Tiime

- **Situation:** Legacy cloud data warehouse costs exceeded €220k/year with slow query execution.
- **Task:** Migrate all pipelines to Snowflake with zero service interruption.
- **Action:** Architected Snowflake schemas and automated ETL transformations.
- **Result:** Cost reduced by >90% (>€200,000/yr saved).

- **Metrics:** >90% Cost Cut (>€200k/yr Saved)

### Ad Traffic Profitability ML & CO2 Reduction (-33%) — Equativ

- **Situation:** High compute energy consumption on ad auctions was squeezing margins and carbon footprint.
- **Task:** Build mathematical bid price & floor optimization algorithms.
- **Action:** Engineered real-time bid floor scoring and filed for CIR research tax credits.
- **Result:** +250% feature MRR and -33% CO2 carbon reduction.

- **Metrics:** +250% Feature MRR & -33% CO2 Carbon Reduction

## Projects

### AEGIS Tactical Radar Sub-App (LIVE SYSTEM)

Real-time AI threat telemetry & technology radar sub-application embedded within this site.

- **Links:** [/radar](/radar)
- **Stack:** SvelteKit 5, TypeScript, Tailwind CSS, Canvas API

### Sudo-Q (LIVE PLATFORM)

Daily Sudoku challenge game to solve in a finite time limit.

- **Links:** [https://sudo-q.com/](https://sudo-q.com/) · [source](https://github.com/Kubyer/sudo-iq.com)
- **Stack:** TypeScript, Sudoku Logic, Daily Challenge, Web App

### SingAndClap Voice AI Benchmark (LIVE BENCHMARK)

Proprietary public evaluation platform to stress-test real-time voice AI models, STT/TTS latency, and audio fidelity.

- **Links:** [https://voice.singandclap.com/](https://voice.singandclap.com/)
- **Stack:** Voice AI, Sub-Second Latency, Python, WebSockets

### Struggl (LIVE APP)

Developer productivity and challenge tracking application.

- **Links:** [https://struggl.singandclap.com/](https://struggl.singandclap.com/)
- **Stack:** Web Application, SaaS, Productivity

## Education

### MSc in Data Analytics & Artificial Intelligence — EDHEC Business School

*Master's Degree*

- **Coursework:** Project Management, Financial Statement Analysis, VBA, Data Analysis, Advanced Statistics, Python, Statistical Models
- **Thesis:** Master Thesis: ML Classifier to recognize drop shipping in the e-commerce industry.

### Classes Préparatoires aux Grandes Écoles (CPGE) — Lycée International de Valbonne

*Pre-Engineering & Math Prep*

- **Coursework:** Mathematics, Computer Science, Physics

## Certifications

- **Dataiku ML Practitioner** — Dataiku (Oct 2024)
  - Skills: Machine Learning, Dataiku DSS

- **Dataiku Core Designer** — Dataiku (Oct 2024)
  - Skills: Data Pipelines, Dataiku DSS

- **Project Initiation: Starting a Successful Project** — Google (Mar 2024)
  - Skills: Product Management, Agile Execution

- **Foundations of Project Management** — Google (Dec 2023)
  - Skills: Product Management

- **Product Academy Series 2023** — McKinsey & Company (Sept 2023)
  - Skills: Product Management, Product Strategy, Roadmapping

- **Google Cloud Big Data & Machine Learning Fundamentals** — Google (May 2022)
  - Skills: Data Analysis, BigQuery ML, GCP Infrastructure

- **Agile au travail : Planifier avec des user stories agiles** — LinkedIn (Feb 2022)
  - Skills: Agile Product Management, User Stories

- **Google Tag Manager Fundamentals** — Google (Oct 2021)
  - Skills: Web Analytics & Tagging

- **Google Analytics Individual Qualification** — Google (Oct 2021)
  - Skills: Data Analysis, Web Analytics

- **Financial Accounting, Financial Statement Analysis & Corporate Finance** — Harvard Business Publishing (May 2020)
  - Skills: Corporate Finance, Financial Analysis