ML/AI Engineer | Cohere Global Negotiation Guide

Negotiation DNA: Pre-IPO Equity + Competitive Base | Enterprise RAG Dominance | Enterprise Sovereignty | +20-35% AGENTIC AI PREMIUM

Region	Base Salary	Equity (Pre-IPO/4yr)	Bonus	Total Comp
Toronto	C$162K-C$208K	$235K-$415K	—	C$228K-C$322K
San Francisco	$195K-$248K	$235K-$415K	—	$254K-$352K
London	£142K-£182K	$235K-$415K	—	£200K-£285K

+20-35% Agentic AI Premium applies for engineers working on agentic RAG, multi-step retrieval agents, and autonomous enterprise AI workflows.

Negotiation DNA ML/AI Engineers at Cohere work on the most critical technical systems in enterprise AI — training and optimizing Command R+ (the #1 enterprise RAG model), building Embed models for semantic search, and tuning Rerank for retrieval quality. You are at the core of what makes Cohere's Enterprise Sovereignty positioning real: building frontier AI models purpose-built for enterprise use cases, optimized for grounded answers from enterprise documents, and deployable within any security perimeter. Founded by Aidan Gomez, co-author of the Transformer paper ("Attention Is All You Need"), Cohere's ML culture combines world-class research with enterprise-grade production requirements — your models must be both state-of-the-art and reliable enough for Fortune 500 regulated deployments.

Level Mapping: Cohere ML/AI Engineer = Google L4-L5 Research Engineer (Toronto) = Meta Research Engineer IC4-IC5 = Anthropic ML Engineer = DeepMind Research Engineer. In the Toronto market — the global epicenter of AI research thanks to the Vector Institute and Geoffrey Hinton's legacy — this role competes with Google Brain Toronto, Meta FAIR, and the growing presence of Anthropic, xAI, and OpenAI research labs. The +20-35% Agentic AI Premium reflects the explosive demand for engineers who can build multi-step retrieval agents and autonomous enterprise AI workflows.

Enterprise Sovereignty — The "Safe" Fortune 500 Alternative Cohere is the "safe" AI choice for Fortune 500 enterprises. While OpenAI and Anthropic are consumer-facing, generate headlines, and raise regulatory concerns, Cohere is enterprise-first, privacy-focused, and deployable within any security perimeter — on-premises, private cloud, or VPC. Fortune 500 CISOs choose Cohere because it's the only frontier AI they can fully control. I build the enterprise AI platform that CISOs trust with their most sensitive data. When negotiating, frame your value as: "I'm not building a chatbot. I'm building the AI platform that Fortune 500 companies trust with regulated data — healthcare, finance, government. Cohere's Enterprise Sovereignty positioning is why we win deals OpenAI can't. My comp should reflect that I'm enabling the highest-value enterprise AI contracts in the market."

Global Levers

1. Agentic AI Premium (20-35%): "Agentic AI — multi-step retrieval agents, autonomous enterprise workflows, and tool-using LLM systems — is the highest-demand ML skill in 2026. I bring direct experience building agentic RAG systems that autonomously retrieve, reason over, and synthesize enterprise documents. This capability is Cohere's next frontier product, and the talent pool is vanishingly small. The 20-35% premium over base ML comp reflects market reality — Google, Anthropic, and OpenAI are all bidding for agentic AI talent with packages 30%+ above standard ML engineer offers."
2. Command R+ Training Expertise: "I contribute directly to training Command R+, the #1 enterprise RAG model. This is not fine-tuning — this is pre-training and RLHF/DPO optimization for enterprise retrieval quality, citation accuracy, and grounded reasoning. The pool of engineers who can improve large-scale model training for enterprise RAG is perhaps 200 people globally. My expertise directly determines whether Cohere maintains its #1 position against GPT-4, Claude, and Gemini in enterprise RAG benchmarks."
3. Pre-IPO ML Engineer Equity: "ML engineers at frontier AI companies command the highest equity grants because our work IS the product. I need $350K+ in equity over four years — at Cohere's $5.5B valuation with a $15-20B+ IPO target, that's $950K-$1.3M at exit. Google Research Engineer in Toronto offers $280K+ in guaranteed public RSUs. Anthropic is offering ML engineers $400K+ in equity at their valuation."
4. Embed/Rerank Model Ownership: "I work on Cohere's Embed and Rerank models — the retrieval infrastructure that makes enterprise RAG work. Embed is the #1 enterprise embedding model for semantic search, and Rerank is the quality layer that ensures retrieval precision. These models serve as the foundation for every enterprise RAG deployment. My model quality improvements directly translate into enterprise customer satisfaction and retention."

Negotiate Up Strategy: "I'm choosing between Cohere, Anthropic, and Google Brain Toronto for my next role. Cohere's Enterprise Sovereignty positioning is uniquely compelling — you're the only frontier AI company that Fortune 500 CISOs can deploy within their security perimeter. Anthropic is offering $230K base with $400K equity over four years. Google Brain Toronto L5 is offering C$215K base with $320K RSUs guaranteed. For Cohere, I'm targeting C$205K base with $380K equity over four years — that includes the Agentic AI premium for my experience building multi-step retrieval agents. At a $15B IPO valuation, that equity is worth $1M+. I also want a research publication agreement — the ability to publish non-proprietary ML research enhances Cohere's brand and my career simultaneously. My accept-at floor is C$185K base with $320K equity — below that, the Anthropic equity upside or Google's guaranteed compensation is the rational choice."

Evidence & Sources

• Levels.fyi — Google L4-L5 Research Engineer, Meta Research Engineer Toronto/SF compensation (2025-2026)
• Blind — Cohere ML Engineer, Anthropic ML Engineer, and frontier AI lab compensation threads (2025-2026)
• PitchBook — Cohere $5.5B+ valuation, Anthropic/OpenAI compensation benchmarks, agentic AI market analysis