How to Negotiate Salary: Scripts, Tactics & What Actually Works
By
Liz Fujiwara
•
Jan 14, 2026
Imagine you’re an AI engineer in 2026 choosing between $260k and $320k offers, a $60k difference that can grow to $700k or more over a decade with raises and equity.
Market rates for roles like LLM development, ML infrastructure, and applied research update every 3–6 months, so initial offers are often below what strong candidates can earn. Scarcity of AI talent gives these professionals significant negotiation power.
Hiring is increasingly AI-assisted, which can make negotiation confusing or clearer depending on design. Fonzi provides transparency by surfacing compensation early and prioritizing skills over pedigree, giving candidates an advantage in negotiations.
Key Takeaways
Most tech companies expect negotiation, and many AI candidates who negotiate receive improved offers, so treat negotiation like an experiment by collecting market data, setting target ranges, and iterating based on recruiter signals.
Modern hiring uses AI, but transparency matters, and platforms like Fonzi use AI to reduce noise and bias while keeping human judgment central.
Concentrated conversations create leverage, as Fonzi’s Match Day compresses multiple high-signal company discussions into a single week, giving candidates clarity on compensation and natural negotiation power.
Understanding Your Market Value as an AI Professional
Before you can negotiate effectively, you need to understand what you are worth in the current market. Many candidates fail here by undervaluing themselves based on previous salaries or anchoring to aspirational numbers without supporting data.
The AI compensation landscape shifted dramatically, with senior roles that paid $200k now regularly exceeding $350k at well-funded companies.
Factors that influence your market rate:
Years of experience with production ML systems (not just research or coursework)
Hands-on experience with LLMs since 2023, including fine-tuning, RLHF, and deployment
Open-source contributions with meaningful adoption
Publications at top venues (NeurIPS, ICML, ACL) for research-oriented roles
Experience at recognized companies or high-growth startups
Geography and remote policy
Tools and resources to benchmark your value:
Levels.fyi for TC benchmarks at FAANG and AI labs (most reliable for established companies)
Glassdoor and Blind for startup compensation and culture insights
2024–2025 AI salary surveys from recruiting firms like Hired and Rora
Pay transparency data from job postings in California, New York, and Colorado
Build a “market value one-pager” summarizing your last three big wins, such as latency reductions, revenue impact from ML models, or training cost optimizations, to reference in negotiation conversations.
Evaluating Your Technical Profile: What You Bring to the Table
Your technical profile determines which salary bands you can realistically target. Different specializations command different premiums:
Applied LLM Engineer (RAG systems, agents, tooling)
High demand, rapidly evolving skill set
Typical TC range: $180k–$350k for mid to senior levels
Premium for production experience with major LLM providers
Infra/DevOps for AI (Kubernetes, GPU orchestration, observability)
Critical for companies scaling inference
Typical TC range: $200k–$380k
Premium for experience with H100/A100 clusters at scale
Research Engineer (model training, evaluation, papers)
Often requires publication track record
Typical TC range: $220k–$450k
Significant premium for NeurIPS/ICML publications
Traditional ML Engineer (recommendation systems, ranking, forecasting)
Mature discipline with established bands
Typical TC range: $170k–$320k
Premium for demonstrated revenue impact
Quantify your impact with specific metrics:
“Reduced inference cost by 35% on A100 clusters”
“Shipped LLM-based feature that increased activation rate by 9%”
“Cut model training time from 72 hours to 18 hours through optimization”
“Improved system reliability from 99.5% to 99.95% for ML APIs”
These numbers become your negotiation ammunition. Vague claims about “improving performance” carry no weight; specific metrics tied to business outcomes justify higher compensation.
Researching Market Rates with Real Data
Complete your market research before any recruiter screening call, whether the role came through Fonzi, LinkedIn, or direct outreach. Walking into early conversations without data puts you at an immediate disadvantage.
Step-by-step research process:
Identify your target title and level, such as Senior ML Engineer, Staff AI Engineer, or Research Scientist.
Gather at least 10 relevant data points from Levels.fyi, filtering by company tier.
Adjust for geography and remote policy using location-specific data.
Define your personal negotiation range: floor as the minimum you would accept without resentment, target as what you genuinely believe you are worth based on data, and stretch as an ambitious but defensible number for strong negotiation positions.
Fonzi pre-normalizes compensation data from partner companies, so candidates see realistic salary ranges upfront rather than guessing, unlike traditional job boards where constraints are often revealed late in the interview process.
When and How to Bring Up Compensation
Timing is critical in salary negotiation. Bring up compensation too early and you anchor yourself before the company is invested; wait too long and you waste time if expectations are misaligned.
The general rule is to share broad salary expectations early for alignment, but save serious negotiation for after a formal offer or near-final verbal confirmation. In typical multi-round processes for AI and infrastructure roles:
Initial screen: Share a range if asked, but focus on fit and interest
Technical interviews: Demonstrate your value; compensation rarely comes up
Final rounds: The company is invested enough that negotiation won’t scare them
Offer stage: This is where real negotiation happens
Salary may appear on postings due to pay-transparency laws, but ranges are often wide and still invite negotiation. A posting showing $180k–$280k indicates the band exists, not that you should accept $180k.
Fonzi’s platform often surfaces an initial range from companies before conversations begin, allowing candidates to respond with data instead of guessing a number.
Early-Stage Conversations: Answering “What Are Your Salary Expectations?”
When recruiters ask about salary requirements in the first ten minutes of a call, they are screening for alignment, not trying to lock you into a number. Your goal is to provide enough information to continue the conversation without anchoring too low.
Script 1: Redirecting with a data-backed range
“I’m looking for market-rate total compensation for senior AI roles, which seems to be around $260k–$340k TC based on recent Levels.fyi data. Does that align with what you’re targeting for this position?”
Script 2: Asking for their range first
“I’m flexible depending on the full compensation package and scope of the role. Could you share the band you’re working with for this position based on the job description?”
Script 3: When you’re currently underpaid
“My current salary isn’t reflective of market rates and I’ve been focused on growth over compensation at my current position. Based on my research, I’m targeting $280k–$320k for my next role.”
Avoid wasting time on roles that can’t meet your floor, but don’t negotiate against yourself by naming a specific number before you understand the full opportunity. Most employers appreciate candidates who can discuss salary with professionalism and data.
When to Avoid Pushing Too Hard
While most tech offers are negotiable, some truly aren’t. Recognizing when to back off preserves relationships and avoids wasting emotional energy.
Signs further negotiation won’t help:
Company explicitly communicates a hard pay band, which is common in government labs or some universities
Equity pool is constrained for legal or board reasons, and this has been explained
They have already matched a strong competing offer and clearly indicated finality
The hiring manager has stated they went to bat for this offer and it is their best
Over-negotiation can sour the relationship before you even start, so optimize for long-term upside such as role fit, learning, and growth instead of last-dollar wins when signals indicate inflexibility.
This doesn’t mean accepting poor offers. It means recognizing when you’ve reached a genuine ceiling and making a clear-eyed decision about whether the opportunity still makes sense.
What Actually Works in Salary Negotiation (Backed by Data and Experience)
Patterns from hundreds of AI and infrastructure negotiations reveal consistent themes: preparation and calm confidence outperform aggression or vagueness every time. Companies expect negotiation from technical candidates, but they do not expect emotional reactivity or unrealistic demands.
Proven tactics that work:
Using specific impact metrics tied to business outcomes
Anchoring with researched ranges from credible sources
Leveraging multiple active conversations, not necessarily offers
Asking open-ended questions to understand flexibility
Being flexible on levers beyond base salary
Framing requests as collaborative problem-solving
A George Mason and Temple University study found that among negotiation styles, competing to maximize personal gain and collaborating for mutual wins yield the highest salary increases, averaging $5,000 above initial offers. Collaborators also reported higher satisfaction with the process, suggesting a blend works best for long-term relationships.
Common Salary Negotiation Tactics for AI Roles
Tactic | When to Use It | How to Phrase It | Risk Level | Best For |
Data-based counteroffer | You have clear market data showing the offer is below median | “Based on Levels.fyi data for L5 AI engineers, median TC is $380k. I’m targeting $360k–$390k.” | Low | Any candidate with solid research |
Leverage competing processes | You have active interviews at comparable companies | “I’m in the final stages with two other companies. I’d love to make a decision by next Friday if we can align on comp.” | Medium | Candidates with 2+ active processes |
Total compensation clarification | Initial offer focused only on base salary | “Can you walk me through the full compensation package including equity, bonus, and benefits?” | Very Low | Everyone (always ask this) |
Trade-offs (equity vs. base) | Base is capped but you believe in company growth | “If base is at the top of band, could we shift $20k into additional equity to reach my target TC?” | Low | Startup/growth-stage candidates |
Non-cash levers | Compensation is firm but you have other priorities | “Would there be flexibility on remote work, learning budget, or conference attendance?” | Very Low | Candidates valuing flexibility |
Signing bonus request | Base and equity are firm but there’s budget for one-time costs | “Is there room for a signing bonus to bridge the gap between the offer and my target?” | Low | Most candidates (easiest lever) |
Exploding-offer pushback | Company gives unreasonably short deadline | “I’m very interested, but I have a final interview on Thursday. Could we extend the deadline to Monday?” | Medium | Candidates with competing processes |
Using Concrete Impact to Justify Your Ask
Companies pay for business impact, not technical curiosity. Your requested compensation becomes easier to justify when tied directly to outcomes you’ve created.
Examples specific to AI and infra roles:
Reducing inference latency from 300ms to 80ms, directly improving user experience metrics
Cutting training costs on H100 clusters by 25 percent, producing quantifiable cost savings
Improving model-driven conversion rate by 12 percent, impacting revenue
Increasing system reliability for LLM APIs from 99.5 percent to 99.95 percent, reducing incidents and customer churn
Negotiation script connecting impact to compensation:
“In my last role, my optimization work on our inference pipeline saved approximately $600k annually in GPU costs while improving p99 latency by 40 percent. Given that level of impact, I am targeting a package that reflects similar value creation, around $340k in total compensation.”
Another example for research-oriented candidates:
“The ranking model I developed drove a 9% increase in click-through rate, which translated to roughly $2M in incremental annual revenue. I’d like the proposed salary to reflect that kind of business contribution.”
Maintain a running “impact doc” throughout your career. Update it quarterly with wins, metrics, and context. When negotiation time comes, you’ll have fresh, quantitative examples ready instead of scrambling to remember details from two years ago.
Scripts and Templates: Exactly What to Say (Email + Live)
Many technical candidates know they should negotiate but freeze when it is time to write the email or have the conversation. This section provides copy-paste-able scripts for real scenarios you will encounter.
These scripts are tailored for AI and infrastructure roles, referencing total compensation, equity, and remote work norms in 2026. They work across US and international markets, though you may need to adjust specific numbers for your geography.
The key principles across all scripts are:
Lead with enthusiasm, not demands
Reference specific data sources
Make a clear ask with a defined range
Leave room for discussion
Script: Countering an Initial Offer via Email
Scenario: You’re a senior AI engineer who received an initial offer of $240k TC. Based on your research and experience, you believe $280k–$300k is appropriate.
Subject: Re: Offer Details – [Your Name]
Hi [Recruiter/Hiring Manager],
Thank you for sending over the offer details. I’m genuinely excited about the opportunity to join [Company] and contribute to [specific project or team].
After reviewing the compensation package, I’d like to discuss the total compensation. Based on my experience deploying LLM-based features that reduced inference costs by 35% and improved user engagement metrics at [Previous Company], combined with current market data from Levels.fyi for senior AI engineers in [location], I’m targeting a TC range of $280k–$300k.
I’m very flexible on how we structure this, whether through base salary adjustments, additional equity, or a signing bonus. My priority is finding a package that reflects the impact I expect to make on [specific team/project].
Would you have time this week to discuss? I’m available [specific times] and happy to jump on a quick call.
Best,
[Your Name]
Variation for Fonzi candidates:
“Based on the ranges I’m seeing across the Fonzi marketplace this quarter for similar roles, and my track record shipping high-reliability ML infrastructure, I believe $290k–$310k TC is appropriate for this position.”
Script: Live Conversation with a Recruiter or Hiring Manager
Conversation flow (3-5 minutes):
Appreciation and enthusiasm (30 seconds)
“Thanks for taking the time to discuss the offer. I’m really excited about this role, and the work on [specific project] aligns perfectly with what I want to focus on next.”Reiterate your fit (30 seconds)
“After going through the interview process and learning more about the challenges you’re solving, I’m confident I can make an immediate impact, especially on [specific area].”Present your range with justification (1 minute)
“Given my experience scaling LLM infrastructure to handle 10M+ daily requests and the market rates I’m seeing for similar roles, I’m targeting closer to $320k–$350k in total compensation. How close can we get to that range?”Pause and listen (let them respond)
Handle pushback with flexibility (1 minute)
“I understand base salary might be at the top of the band. Is there flexibility on equity or a signing bonus to bridge the gap? I’m also open to discussing a performance-based compensation review after 6-12 months.”
Key phrases to memorize:
“Based on my impact at [Company], I’m targeting…”
“I’m flexible on structure; base, equity, or signing bonus all work for me.”
“What would it take to get closer to [target number]?”
Script: Negotiating When You Have No Other Offers
Many candidates worry they have no leverage without competing offers. This isn’t true because your leverage comes from your value, not external pressure.
Email script:
Hi [Recruiter],
Thank you again for the offer! I’m very excited about joining [Company] and contributing to [team/project].
I want to be transparent: [Company] is my top choice, and I’m not using other offers to create artificial pressure. That said, based on current market data for AI engineers with my background in [specific area] and my track record [specific achievement], I believe a TC closer to $270k–$290k would be appropriate.
I’m confident I can deliver significant value in this role, particularly on [specific challenges they’re facing]. Would there be room to adjust the package to better reflect that expected impact?
Happy to discuss further whenever it works for you.
Best,
[Your Name]
Key points when you lack competing offers:
Emphasize your unique value and track record
Reference market data, as employers expect this
Focus on future contribution potential
Make a modest but meaningful ask, typically 5-10 percent above the initial offer
It is okay to negotiate even without competing offers, as employers expect this from specialized AI talent and reasonable requests rarely result in rescinded offers
Beyond Base Salary: Total Compensation and Non-Salary Levers
For AI and ML roles, base salary is often just 40-60% of total compensation. Fixating only on base means ignoring the majority of your potential earnings.
Typical TC structures by company stage:
Company Stage | Base Salary | Equity | Signing Bonus | Performance Bonus |
Early-stage startup | 60-70% of TC | 25-35% (higher risk/reward) | 0-5% | Rare |
Growth-stage (Series B-D) | 50-60% of TC | 30-40% | 5-10% | 5-10% |
Large tech (FAANG, etc.) | 40-50% of TC | 35-45% | 5-15% | 10-15% |
Non-salary levers particularly relevant for AI work:
Remote versus hybrid policy, which has significant lifestyle and cost-of-living impact
Conference and compute budgets, critical for staying current
Time for open-source contributions, important for career building
On-call expectations and rotation frequency
Internal mobility options, allowing movement between teams or projects
Learning stipends and education benefits
Create a personal priority ranking before negotiating. If full remote matters more than an extra $10k in base, you can make strategic trades. If equity upside excites you more than cash stability, you can negotiate accordingly.
Trading Base Salary for Equity, Bonus, or Perks
Sometimes base salary is genuinely capped, but other elements have flexibility. Knowing how to make trades unlocks additional value.
Example 1: Equity trade at a growth-stage startup
A Series B AI startup offers $200k base + $150k equity (over 4 years) + $30k signing bonus = $242.5k annualized TC. You want $270k.
“I understand base salary is at the top of your band. Would you consider an additional 0.05% in equity and increasing the signing bonus to $50k? That would get us closer to my target and align my incentives with long-term company success.”
Example 2: Remote work trade
A company offers $280k TC but requires Bay Area office presence. You want to work remotely from a lower cost-of-living city.
“I’m very interested in this role, and I understand the value of in-person collaboration. Would there be flexibility for fully remote work if I committed to quarterly on-sites? I’m willing to accept the current compensation without adjustment in exchange for that flexibility.”
Evaluating startup equity:
Understand vesting schedules (typically 4 years with 1-year cliff)
Research recent funding rounds and implied valuation
Consider realistic exit scenarios (acquisition more common than IPO)
Ask about strike price and current 409A valuation
Factor in dilution from future funding rounds
Negotiating Role Scope, Title, and Growth Path
For AI engineers, title and scope can heavily influence future earnings. The difference between Senior and Staff roles often represents $50k-$100k in annual compensation at the same company, and owning a critical system versus contributing to it affects your leverage in future negotiations.
What to negotiate beyond compensation:
Title clarity: Ensure your level matches industry standards (L5 vs. Senior, L6 vs. Staff)
Scope ownership: Define specific areas you’ll own (e.g., “leading LLM evaluation infrastructure”)
Growth timeline: Request a written 6-12 month promotion and comp review plan
Team placement: Negotiate for high-impact projects aligned with your career goals
Sample script for scope negotiation:
“I’m willing to accept the current compensation if we can document a clear path to Staff level within 12-18 months, contingent on delivering [specific outcomes]. Can we include that expectation in my offer letter?”
Companies that clearly articulate role scope and growth paths demonstrate maturity in both technical and people practices. View this transparency as a green flag when evaluating offers.
How AI Is Changing Hiring—and How Fonzi Uses It Differently
Most companies now use AI somewhere in their hiring process, including resume screening, candidate ranking, automated scheduling, and sometimes AI-generated outreach. For technical talent, this creates a strange dynamic because you build AI systems but are also being evaluated by them.
Common AI hiring tools and their limitations:
Keyword-based resume parsers that miss context and favor buzzword-stuffed resumes
Automated coding-screen scoring that may not reflect real-world engineering judgment
AI-generated recruiter outreach that feels impersonal and irrelevant
Predictive “fit” models often trained on historical data that encodes bias
The risks are real: these systems can reinforce pedigree over skills, disadvantage non-traditional or self-taught engineers, and make it hard to understand why you’re being rejected or advanced.
Fonzi takes a different approach: use AI to create clarity, not confusion. The platform matches candidates and companies based on skills, preferences, and portfolio work while keeping humans at the center of final decisions.
How Fonzi’s Matching Works for AI and ML Talent
Fonzi operates as a curated marketplace specifically for AI, ML, and infrastructure talent. Instead of applying to hundreds of roles and hoping keywords align, candidates are vetted for relevant skills and matched with companies actively hiring for those exact profiles.
How the matching works:
Candidates submit profiles including GitHub activity, past roles, project descriptions, and specific technical interests
Fonzi’s AI processes these signals to suggest relevant matches, such as companies looking for LLM agent experience, real-time inference infrastructure, or applied research
Human reviewers validate matches to catch algorithmic blind spots and avoid bias
Candidates see transparent information about companies before engaging, including tech stack, compensation bands, role scope, and location expectations
This contrasts sharply with traditional job boards where you apply into a black box. On Fonzi, you know the salary information and role constraints upfront, giving you better negotiation footing from day one.
Reducing Bias and Protecting Candidate Experience
Poorly designed AI tools encode bias against non-traditional paths, including bootcamp graduates, self-taught engineers, and career-switchers from adjacent fields. This is particularly problematic in emerging areas like LLM engineering, where formal credentials are rare.
Fonzi’s bias-reduction practices:
Focus on skills and demonstrable work, such as repos, demos, and research, over school names
Monitor matching models for biased recommendations
Give candidates control over what information is shared with companies
Prioritize response time transparency and feedback loops
Consider a self-taught infra engineer who built large-scale GPU clusters at a smaller company. Traditional resume screens might filter them out for lacking a CS degree from a top-20 school. Fonzi’s skills-based matching surfaces their actual capabilities.
Communication on Fonzi, especially around compensation, is designed to be respectful and human-driven, providing expected response times and actual feedback rather than silence.
Match Day: A High-Signal, Efficient Way to Attract Offers
Match Day is a recurring event where pre-vetted AI and infra candidates go live to high-quality, actively hiring companies in a concentrated window.
How Match Day works:
Candidates are vetted and matched based on skills and preferences
On Match Day (typically weekly or biweekly), matched companies receive access to candidate profiles
Companies reach out rapidly during the concentrated window
Candidates can line up multiple processes within 1-2 weeks
Why this matters for negotiation:
Natural leverage from multiple conversations happening simultaneously without coordinating dozens of separate job boards
Reduced exhaustion with fewer cold applications and more focused interviews
Realistic ranges early as companies share compensation expectations upfront, respecting everyone’s time
Parallel options, where having 3-4 active conversations gives honest leverage without manufacturing competing offers
Match Day compresses what traditionally takes months into weeks, giving you clarity and momentum when it matters most.
Preparing for Interviews and Showcasing Your Skills
The strongest negotiation position comes from genuine company excitement about your fit, which is earned through strong interviews and compelling work samples.
Preparation best practices for AI professionals:
Review system design for ML pipelines, including feature stores, model serving, and monitoring
Brush up on LLM architecture basics, such as transformers, attention mechanisms, and RLHF
Study infrastructure scaling patterns, including Kubernetes, GPU orchestration, and observability
Practice coding in Python with an emphasis on clarity and edge case handling
Assemble a concise portfolio:
GitHub repos with clean, documented code
Short write-ups of impactful projects (1-2 pages)
Kaggle or similar competition results if relevant
Links to talks, blog posts, or technical writing
The better you demonstrate impact and clarity of thought in interviews, the easier it is to justify higher compensation during the negotiation phase.
Technical Interview Prep for AI, ML, and Infra Roles
Interview formats in 2026 vary significantly by company and role type:
Interview Type | What They Test | How to Prepare |
Take-home LLM tasks | Practical implementation, prompt engineering, evaluation | Practice building small RAG systems or agent workflows |
Pair programming | Real-time coding, collaboration, communication | LeetCode for pattern recognition, then practice explaining your thinking aloud |
System design (ML) | Architecture thinking, trade-offs, scalability | Study feature stores, model serving patterns, A/B testing infrastructure |
Model evaluation deep dives | Statistical reasoning, metric selection, failure analysis | Review your past work and be ready to discuss evaluation decisions in detail |
Research interviews | Novel thinking, paper familiarity, methodology rigor | Know your publications deeply, and be ready to critique your own methods |
Resources worth your time:
LeetCode or HackerRank for coding pattern familiarity
“Designing Machine Learning Systems” by Chip Huyen for ML systems thinking
Recent papers on your specific focus area (LLMs, infra, ranking, etc.)
Mock interviews with peers who can give honest feedback
Fonzi may provide preparation support to help candidates enter Match Day confident and ready to perform.
Presenting Projects and Portfolios That Strengthen Your Negotiation Position
Curate 2-3 flagship projects that align tightly with your target roles:
For each project, be ready to discuss:
Problem context: What business or technical problem were you solving?
Constraints: What were the latency, cost, reliability, or scale requirements?
Your specific contributions: What did you personally design, build, or improve?
Results: What metrics moved? By how much?
Retrospective: What would you do differently with more time or resources?
Example project presentation:
“At [Company], I led the redesign of our LLM-based summarization pipeline. The original system had 2.3-second latency and cost $0.04 per request. I implemented a smaller distilled model with speculative decoding, reducing latency to 400ms and cost to $0.008 per request, achieving an 80 percent cost reduction while keeping quality scores within 2 percent of the original.”
Upload or link these projects in your Fonzi profile so companies see them before interviews. Concrete, visible work products often carry more weight than resumes alone, especially among engineering-driven companies where hiring manager decisions depend on demonstrated capability.
Conclusion
Salary negotiation is not about winning against the employer; it is about finding alignment between your value and their investment in you. The skills you have developed as an AI engineer, ML researcher, infrastructure engineer, or LLM specialist are genuinely scarce and valuable. Negotiation simply ensures that value is reflected in your compensation.
In 2026, AI is embedded throughout the hiring process. The question is whether that AI amplifies human judgment or replaces it, and whether it creates clarity or confusion. Fonzi optimizes explicitly for a candidate-first approach, with transparent compensation data, skills-based matching, and humans at the center of final decisions.
Adopt a calm, data-driven mindset. Know your market value. Prepare specific impact stories. Treat compensation as a holistic package rather than just your salary. Remember that negotiation is a professional conversation between equals, not a favor being requested.
