Code-review Empregos à distância e no escritório em casa ∙ Página 2

About the Role

We’re hiring a Staff Engineer – Core AI to design, experiment, and scale the next generation of LLM-powered multi-agent systems that enable intelligent, secure, and compliant automation for financial institutions. This role goes beyond integrating third-party APIs — it’s about building differentiated intelligence: training, tuning, and evolving models that reason, plan, and act autonomously in high-stakes environments. You’ll work at the intersection of LLM research, applied reinforcement learning, and AI systems engineering, driving innovation in model fine-tuning, prompt optimization, encryption for inference, and speech-to-speech AI.

Your mission: create the AI runtime layer that powers adaptive, explainable, and policy-aligned agents — at scale.

What You’ll Own 

As the lead for LLM engineering, you’ll define how models learn, optimize, and safely interact with sensitive financial data. You’ll be responsible for:

• Model Evolution: Building fine-tuning pipelines, exploring open-weight models, and benchmarking their performance against proprietary LLMs.
• Inference Optimization: Driving high-throughput, low-latency inference strategies across GPUs, TPUs, and distributed inference clusters.
• Safety & Guardrails: Designing data-safe pipelines with encryption for model I/O, and implementing automated PII detection and masking at both prompt and response layers.
• RL-Based Learning: Applying Reinforcement Learning (RLHF/RLAIF), reward modeling, and policy optimization to continuously improve model performance.
• Speech-to-Speech and Multimodal AI: Exploring speech model architectures (ASR/TTS) and building adaptive pipelines for natural, real-time conversational intelligence.
• POCs & Experimentation: Rapidly prototyping emerging models, toolchains, and optimization methods to maintain a competitive edge.
• Framework Leadership: Collaborating with research and backend teams to evolve our custom AI orchestration layer — combining multiple specialized models, memory systems, and evaluation tools.

What You’ll Do

• Lead Fine-Tuning and Experimentation: Create fine-tuning workflows using LoRA, PEFT, and instruction-tuning pipelines; manage large-scale training datasets.
• Drive Auto-Prompt Optimization: Build self-evolving prompt evaluation loops using reinforcement learning, reward modeling, and continuous evaluation frameworks.
• Accelerate Inference Throughput: Optimize model inference through quantization, batching, caching, and high-performance serving strategies.
• Implement Encrypted Inference: Develop novel encryption and key management techniques for model-level data protection during inferencing.
• Design Guardrail Systems: Implement policy layers that enforce safety, prevent hallucinations, and ensure compliance (SOC2, GDPR).
• Integrate Speech Models: Develop and optimize speech-to-speech pipelines, managing end-to-end latency, transcription accuracy, and model adaptation.
• Run Advanced Evals: Establish evaluation harnesses that measure factual accuracy, latency, cost-efficiency, and safety compliance in production environments.
• Research and Publish: Explore the latest advancements in open-source LLMs and reinforcement learning for agents, driving our internal AI innovation roadmap.

What We’re Looking For

Required Qualifications

• Strong LLM Expertise: 2+ years of experience working directly with transformer architectures and LLM fine-tuning (e.g., Llama, Mistral, GPT, Mixtral, Gemma, Falcon, Claude)
• Applied Reinforcement Learning: Hands-on experience with RLHF/RLAIF, reward modeling, and multi-objective optimization for generative models
• Prompt Optimization & Evaluation: Deep knowledge of auto-prompting, chain-of-thought evaluation, and self-improving agent loops.
• Inference Engineering: Experience improving throughput, quantization, and token efficiency on GPUs or specialized inference hardware.
• Data Security in AI: Knowledge of PII masking, data encryption, and secure model pipelines in production settings.
• Modern AI Tooling: Experience with frameworks such as PyTorch, Transformers, Deep Speed, Hugging Face, LangChain, or vLLM.

Preferred Qualifications

• Experience with speech-to-speech or multimodal models (ASR, TTS, embeddings)
• Understanding of AI evaluation frameworks (e.g., Evals, Llama Index Benchmarks, or custom metrics)
• Familiarity with financial data compliance and AI observability tools
• Exposure to low-level inference optimization (CUDA kernels, model parallelism).
• Contributions to open-source LLM or RL research projects

What Makes This Role Special?

• You’ll shape the core AI that powers agentic intelligence for financial systems serving millions of users.
• You’ll own a research-meets-engineering mandate — from exploring new models to bringing them to life in production.
• You’ll define how autonomous AI systems learn, adapt, and remain safe in a regulated environment.
• You’ll work with a team combining AI research, applied data science, and product engineering, moving fast with purpose and rigor.

Compensation

•  Compensation is expected to be between $200,000 - $240,000. Exact compensation may vary based on skills and location.

What We Offer

• 💡 100% paid health, dental & vision care
• 💰 401(k) match & financial wellness perks
• 🌴 Discretionary PTO + paid parental leave
• 🧠 Mental health, wellness & family benefits
• 🚀 A mission-driven team shaping the future of banking