Adnane Aboutalib

(HEAD -> teal-technology-services) Feb 2026 — Present

work: AI Engineer at Teal Technology Services

Casablanca, Morocco

• Built a multi-tenant GenAI SaaS for market and competitive intelligence, turning web, news and social ingestion into strategic insights, alerts and early weak-signal detection.
• Designed LangGraph pipelines for multi-provider ingestion, semantic deduplication, LLM relevance judging, summarization and 1-5 impact-severity classification, with OpenAI and Voyage AI embeddings.
• Architected weak-signal detection as a LangGraph state graph: pgvector ANN matching, multi-factor promotion scoring, LLM-as-judge adjudication and pooled-evidence backfill for cross-topic discovery.
• Shipped Dockerized FastAPI/SQLAlchemy services with RabbitMQ retry/dead-letter routing, Postgres-checkpointed crash recovery, multi-tenant pgvector search, Redis caching with pub/sub SSE, and LangSmith tracing.

Technologies: Python, LangGraph, LangSmith, OpenAI, Voyage AI, FastAPI, SQLAlchemy, PostgreSQL, pgvector, RabbitMQ, Redis, Docker

Sep 2025 — Feb 2026

work: AI / Data Engineer at ELAVI

Casablanca, Morocco

• Built an AI phone agent that searches vector databases, books calls through calendars, uses external tools and transfers calls.
• Integrated the agent with telephony providers and built Flask APIs for usage analytics.
• Cut manual workload by 40% and automated 17 daily calls and tasks.
• Analysed calls to surface customer pain points and agent pitfalls.

Technologies: Python, Flask, LLMs, Vector databases, Telephony APIs

(vidyze) May 2025

build: vidyze.co

Chapter markers for long YouTube tutorials: 2.8x faster responses, 40% less memory, 5.3x cheaper to run.

Jul 2019 — Jul 2024

edu: Master of Engineering, Data Engineering at National School of Applied Sciences (ENSA) Al Hoceima

Al Hoceima, Morocco

Feb 2024 — Aug 2024

work: Machine Learning Engineer Intern at Digiole Oy

Helsinki, Finland

• Built the GraphQL API Back-end Creator, which generates documented GraphQL schemas and back-ends on AWS.
• Used an LLM to turn schemas in several formats into GraphQL schemas with inline documentation.
• Provisioned every resource as Infrastructure as Code.

Technologies: Python, LLMs, GraphQL, AWS, Infrastructure as Code

(youtube-views-prediction) Nov 2023

build: youtube-views-prediction

End-to-end ML platform: YouTube API data in MongoDB, CatBoost/XGBoost models, deployed with Docker on Azure.

(emailgetter) May 2023

build: emailgetter.co

Influencer outreach optimisation for brands.

(hult-prize-accelerator) May 2021

build: Hult Prize accelerator

Took Neoshelf, my AI inventory startup, to the 2021 global accelerator in London: one of 50+ teams from 300,000+ applicants.