Staff AI Engineer

Staff AI Engineer 

About Akoncagua AI 

Akoncagua AI is building domain-specialized foundation models for technically demanding industries. Our work spans large language models, multimodal systems, synthetic data generation, alignment, evaluation, and scalable training infrastructure 

We are seeking an exceptional Staff AI Engineer who combines deep machine learning expertise with strong software engineering fundamentals to lead our model development efforts. This role is ideal for individuals who can provide the technical direction of foundation model work while remaining deeply hands-on. You will set the architecture and research agenda across the entire model development lifecycle—from experimentation and data curation to training, evaluation, optimization, and deployment—and lead a team of researchers and engineers in executing it. As a Staff AI Engineer, you'll make the core modeling and infrastructure decisions, mentor the team through design reviews and hands-on collaboration, and serve as the technical anchor for building state-of-the-art language and multimodal AI systems. 

What You'll Do 

• Provide the technical vision and roadmap for our domain VLM, translating ambitious long-term goals into concrete quarterly milestones and shippable model releases. 

• Lead the ML team day-to-day: set priorities, assign and sequence work, run technical planning, and own delivery of model milestones. 

• Mentor and grow engineers technically through design reviews, pairing, and direct, honest feedback. 

• Stay hands-on: design and run training experiments (pretraining, SFT, long-context extension), write reusable training/eval infrastructure, and personally dig into model failures. 

• Make core architecture and modeling decisions—model selection, data strategy, multimodal fusion, context-length extension, evaluation methodology—and defend them against primary sources rather than inherited assumptions. 

•Provide the path from research to production: training infrastructure, distributed/long-context training, inference optimization, and deployment reliability. 

•Establish rigorous experimental practice—clean baselines, reproducible evals, honest reporting of results. 

• Partner cross-functionally with product, data, and domain experts to keep the model grounded in real user needs. 

• Make build-vs-adopt calls on the fast-moving open-source and research landscape (new base models, training frameworks, compression and retrieval techniques). 

Minimum Qualifications 

• 8 years of ML/software engineering experience, with significant time spent training or fine-tuning large multimodal models 

• Deep hands-on experience training large vision-language or language models and deep understanding on their architectures. 

• Strong fluency in Python and PyTorch, and comfort with distributed training. 

•Experience taking models from research, prototype to production. 

•Data engineering for training at scale: large-corpus curation, deduplication, quality filtering, tokenization pipelines, data mixtures, and sequence packing. 

Preferred Qualifications 

• Experience leading a small engineering team as a tech lead for setting direction, coordinating delivery, and mentoring. 

• Experience with long-context training and context-extension methods 

• Familiarity with modern training and serving stacks. 

• Track record of shipping applied ML systems in a specific domain, or first-author publications at top venues (NeurIPS, CVPR, ICML, ICLR, ICCV, ACL) 

• Solid grasp of scaling laws and compute-/data-optimal trade-offs (model size vs. data vs. compute) and the ability to plan training runs and budgets against them. 

• Post-training / alignment fundamentals: supervised fine-tuning (SFT) and preference optimization (RLHF, DPO, or similar). 

• Familiarity with AI safety, alignment, and responsible-scaling practices in a production training context. 

Pay Range: $120,000.00 to $160,000.00