---
title: "Research Scientist, Program Synthesis & Code Reasoning"
slug: "research-scientist-program-synthesis-code"
id: "8e100787-891f-46e1-a0b7-89ae7bfeb028"
department: "AI Research"
department_code: "RES"
location: "Paris, France"
workplace_type: "Hybrid"
employment_type: "Full-time"
compensation: "€125k – €175k (annual package)"
is_remote: false
is_featured: false
headcount: 1
canonical_url: "https://jobs.talent247.co/jobs/research-scientist-program-synthesis-code"
markdown_url: "https://jobs.talent247.co/jobs/research-scientist-program-synthesis-code.md"
apply_endpoint: "https://jobs.talent247.co/api/public/jobs/research-scientist-program-synthesis-code/apply"
schema_endpoint: "https://jobs.talent247.co/api/public/jobs/research-scientist-program-synthesis-code/form-schema"
webmcp_tool: "apply_for_role"
updated_at: "2026-11-15T17:25:15.297Z"
content_signal: "search=yes, ai-input=yes, ai-train=no"
---

# Research Scientist, Program Synthesis & Code Reasoning

## Position Telemetry

- **Reference:** `RESEARCH-SCIENTIST`
- **Practice / Sector:** AI Research (`RES`)
- **Location:** Paris, France
- **Workplace Type:** Hybrid
- **Employment Type:** Full-time
- **Compensation:** €125k – €175k (annual package)
- **Canonical HTML URL:** https://jobs.talent247.co/jobs/research-scientist-program-synthesis-code

## Role Dossier & Specification

THE CLIENT & MANDATE
Our client is a $500M+-capitalized frontier AI laboratory operating its flagship R&D headquarters in Paris, purpose-built to solve autonomous software engineering from first principles. Rather than treating code generation merely as next-token text prediction over public repositories, they pre-train and post-train proprietary foundation models closed-loop against millions of live compiler, static-analyzer, and execution environments—deploying their models directly inside security-conscious global enterprises and defence-industrial organizations.

They have engaged Talent247 to appoint a Research Scientist, Program Synthesis & Code Reasoning in Paris. Working alongside world-class RL researchers, programming-language theorists, and supercomputing systems engineers, you will invent new training and test-time search paradigms that enable foundation models to reason soundly across million-line repositories, formal invariants, and complex multi-step architecture changes.

KEY RESPONSIBILITIES
• Pioneer reinforcement learning and test-time search algorithms (MCTS, verifier-guided beam/tree search, process reward models, RLVR) grounded in compiler diagnostics, type checkers, fuzzers, and unit/integration test execution.
• Design synthetic repository evolution and bug-injection curricula that teach long-horizon repository navigation, cross-file refactoring, root-cause debugging, and security-vulnerability patching.
• Explore neuro-symbolic bridges combining formal verification (Lean, Dafny, SMT solvers, abstract interpretation) with large-scale autoregressive and diffusion code models.
• Run rigorous scaling-law ablations across data mixtures, context lengths (128k–1M+ tokens), and RL compute allocation on dedicated multi-thousand-GPU clusters.
• Collaborate closely with distributed execution-sandbox engineers to ensure reward signals remain fast, deterministic, and immune to reward hacking or test tampering.
• Translate breakthrough research directly into flagship enterprise model releases.

TECHNICAL ENVIRONMENT & OFFERING
Researchers work on a massive dedicated H100/H200 GPU cluster with custom PyTorch/JAX distributed training frameworks and a planet-scale sandboxed code-execution farm. Based hybrid in Paris (with relocation support across Europe and globally), our client offers €125,000–€175,000 EUR base salary plus top-decile frontier-lab equity, private healthcare, and full French Talent Passport visa sponsorship.

## Core Requirements

ESSENTIAL QUALIFICATIONS
• PhD in Machine Learning, Programming Languages, Formal Methods, or Computer Science, or equivalent frontier lab track record in LLM reasoning and code synthesis.
• Publications at premier venues (NeurIPS, ICML, ICLR, POPL, PLDI, OOPSLA, CAV, or ICSE) or demonstrated leadership on frontier code/reasoning foundation models.
• Strong hands-on proficiency in PyTorch or JAX for large-scale LLM pre-training, post-training (RLHF/GRPO/PPO), or neural search algorithms.
• Deep familiarity with static program analysis, AST/CFG/SSA representations, type systems, compilers, or automated theorem proving / fuzzing.
• Ability to design clean controlled experiments and collaborate tightly with systems engineers at thousand-GPU scale.

PREFERRED / BONUS EXPERIENCE
• Direct experience mitigating reward hacking in verifiable code or math RL environments (e.g., SWE-bench, LiveCodeBench, repository-scale benchmarks).
• Systems fluency in Rust or C++ alongside Python.

## Application Schema & Direct Apply (`apply_for_role`)

Candidates and authorized AI agents on behalf of a candidate can submit an application via the declarative WebMCP form tool `apply_for_role` on https://jobs.talent247.co/jobs/research-scientist-program-synthesis-code or via HTTP `POST https://jobs.talent247.co/api/public/jobs/research-scientist-program-synthesis-code/apply` (`application/json` or `multipart/form-data`). Non-browser HTTP agents must include header `X-Talent247-Intake-Token` obtained from `GET https://jobs.talent247.co/api/public/intake-token` (or automatically returned in `GET https://jobs.talent247.co/api/public/jobs/research-scientist-program-synthesis-code/form-schema`).

### Form Fields

- `full_name` (text, required): Full name
- `email` (email, required): Email
- `phone` (phone, optional): Phone
- `location` (text, optional): Location
- `linkedin_url` (url, optional): LinkedIn URL
- `cover_letter` (textarea, optional): Cover letter
- `consent` (checkbox, required): I consent to my data being processed for this application
- `cv_file` / `cv_base64` + `cv_filename`: Upload a CV in .pdf or .docx format (<= 10 MB) or provide an `https://` LinkedIn profile URL in `linkedin_url`.

### Alternative: Talent Pool Intake (`join_talent_pool`)

If this mandate is not an exact match, register once in the Talent247 Talent Pool at https://jobs.talent247.co/talent-pool (`https://jobs.talent247.co/talent-pool.md`) via WebMCP tool `join_talent_pool` or `POST https://jobs.talent247.co/api/public/talent-pool`.
