Data Scientist / AI Engineer (NLP & LLM)

Omniscient; Paris; On-site; EUR 55,000 - 85,000; Full time

Job summary

Omniscient, based in Paris, is seeking a data scientist to take NLP and LLM work from experiments to production. You will shape how multilingual signals are cleaned, classified, scored and enriched, turning data into actionable intelligence for executives.

The role sits at the intersection of research, product and engineering, with a focus on building reliable, scalable pipelines, coding agents, and shipping end-to-end solutions on-site in Paris three days a week.

Qualifications

• Hands-on experience with modern LLM-based approaches.
• Production-grade Python/SQL coding experience.
• At least two years of professional experience after graduation.
• PhD or research experience in NLP/ML is a plus.
• Experience shipping models or LLM features.
• Willingness to work on-site in Paris three days a week.

Responsibilities

• Own NLP/ML projects from prototype to production.
• Shape multilingual signal cleaning, classification, scoring and enrichment.
• Build and ship end-to-end solutions across data, models and UX.
• Design and enhance the multi-agent layer and data integrations.
• Monitor models, evaluate cost and ensure reliability.
• Collaborate across research, product and engineering teams.

Skills

Applied NLP and ML

Python

SQL

Shipping ML models / LLM features

Fluent English

Education

PhD or research experience in NLP/ML

Tools

Claude Code

Codex

Cursor

Job description

Omniscient delivers real-time intelligence that helps organisations and leaders make better decisions, faster.

Our platform brings together public and internal data to help corporates, investors and public institutions detect risks, understand shifts in their environment and act with greater confidence.

Combining intuitive dashboards with AI-powered analysis, Omniscient turns complex, fast-moving signals into clear, actionable intelligence, from reputation monitoring and geopolitical developments to competitive positioning and narrative change.

We help decision-makers track what matters across their ecosystem, benchmark against peers, and stay ahead of emerging threats and opportunities.

The role

We are hiring a data scientist who ships: someone who takes NLP and LLM work from experiment to production, and owns it there.

You will shape how Omniscient reads the world: how large volumes of multilingual signals are cleaned, classified, scored and enriched, and how our agents turn them into analysis an executive can act on. The role sits where research, product and engineering meet.

The role is broad by design. Models, agents and data are the core, but a large share of the work is the full-stack and infrastructure that bring them to our users. We are a small team: when the fire is elsewhere, you go there. We need someone who builds reliably, thinks in systems, uses coding agents effectively, and helps the team ship faster without compromising quality.

What you will do

• Product mindset. Understand what an executive or an analyst does with what the models produce. Turn a business need into a task you can measure. Own part of the PO or PM duties on the AI side.
• Models and scoring. Classification, sentiment, entity recognition and resolution, topic extraction, summarisation, translation and risk scoring, on multilingual text and increasingly on images and video. Pick the right approach for each job: prompting, fine-tuning, embeddings or classic ML.
• Agents. Design and improve our multi-agent layer: tools, context, orchestration. Connect new data sources and tools to the agents, including through MCP.
• Retrieval. Embeddings, chunking, indexing and retrieval quality. Give the agents context they can actually use.
• Evaluation. Build evaluation sets and harnesses. Measure quality, drift, hallucination and cost per call. We want to get to the point where nothing ships without a number, and you will lead that.
• Data. Explore, clean and label large, messy text datasets. Deduplicate, normalise, find the gaps. Work hand in hand with data engineering on pipelines and data quality.
• Full-stack development. Own the UX, the back end and all the supporting infrastructure that bring an agent or a model to life for our users.
• Research. Follow the literature, run experiments, write them up so the team can build on them. Work with academic partners when it helps.
• Cost. LLM spend is a first-class design constraint: the right model for each job, caching, batching, distillation.
• Production. Ship models and agents as services (C# and Python), monitor them, own them. You build it, you run it. You take your turn on call.
• Productivity. Use coding agents wisely. Write skills and routines. Automate what can be automated. Constantly strive to increase the productivity of the team.

What we look for

• Applied NLP and ML. Strong fundamentals, with hands-on experience of modern LLM-based approaches: transformers, embeddings, fine-tuning, evaluation, retrieval-augmented workflows.
• Python and SQL. You write production code, not only notebooks, and a statically typed language does not scare you.
• Shipped and lived with it. At least one model or LLM feature taken into production and maintained afterwards.
• Research depth, a real plus. A PhD or research experience in NLP, machine learning, computational linguistics or applied mathematics, as long as you want to ship.
• Experience. At least two years of professional experience after graduation.
• A builder at heart. You have a genuine passion for shipping slick products. Here, that means crafting a new form of executive intelligence.
• Agentic development. We are an AI-powered organisation, laser-focused and built around efficiency and agentic workflows. You already build with coding agents (Claude Code, Codex, Cursor or similar) and they are central to how you work. You keep learning, look for better ways, automate your own work and share with your peers. We want people with the same drive to build fast and efficiently.
• Fundamentals over framework years. With agents, we need fewer specialists in one tool and more people with sharp concepts and the appetite to learn. We do not need eight years in each of our frameworks.
• Fluent professional English. You work with colleagues and clients in English without friction. French is a plus, not a requirement.
• Paris. You live in or around Paris, within commuting distance of the office, and work on site at least three days a week.

We know it’s a lot, and we don’t expect you to know everything: equivalent technologies and experiences count too. But learn, be afraid of nothing, collaborate, and grow. And Claude is here to help.

Desired character traits:

autonomy, team player, curious, fast learner, committed.

How we work

• Make it happen. We are a startup. When something needs doing, we do it and learn what we need on the way. Everyone has a spike, everyone works across the stack, and “not my speciality” is not something we say.
• You build it, you run it. There is no separate ops team. Each of us owns observability, alerts and incidents for what we ship, and we share an on-call rotation across the working week. Our clients rely on Omniscient when they face a crisis: being down at that moment is not an option.
• No single point of failure. Every critical system is known by at least two people, with the access to match.
• Agents everywhere. We build with coding agents every day. We expect you to multiply yourself with them: write skills and routines, run loops, industrialise what works, and share it with the team.

The essentials

• Location: Paris, in the office at least three days a week (candidates must live within commuting distance)
• Experience: at least two years after graduation
• Contract: permanent (CDI), full-time
• Languages: fluent English required, French a plus