AI engineering comp data, sourced from offers.
5,800+ network-verified offers across 12 frontier labs and 40+ AI-native companies. By role, by level, by lab, by geography. Updated monthly. The most comprehensive AI engineering compensation dataset in 2026 -- and the only one built on offers we've actually seen.
Network-verified offers in the May 2026 dataset.
Companies tracked -- 12 frontier labs + 40 AI-native.
Specialized role tracks, across research, ML eng, and applied AI.
US, UK, EU, India, Singapore, Canada, MENA, LATAM cuts.
// The master table
Senior IC total comp by role -- May 2026.
Median total compensation (base + equity + bonus, annualized USD) for senior IC offers with 5–8 years of experience. US-based unless noted. Filter by track above.
Senior IC offers -- 24 role tracks
Click any role to see lab-by-lab breakdown in your account (login required).
| Role | Track | Range (USD) | Median | YoY |
|---|---|---|---|---|
| Pre-training research | Research | $700K–$1.4M | $960K | +14% |
| Post-training research | Research | $640K–$1.2M | $820K | +18% |
| Alignment & safety | Research | $680K–$1.1M | $790K | +22% |
| Interpretability | Research | $650K–$1.0M | $760K | +11% |
| Scaling research | Research | $720K–$1.3M | $910K | +9% |
| Agent / RL research | Research | $660K–$1.15M | $820K | +24% |
| Multimodal research | Research | $650K–$1.1M | $780K | +13% |
| Evals research | Research | $580K–$950K | $710K | +15% |
| Training platform engineering | ML eng | $700K–$1.35M | $920K | +12% |
| Inference engineering | ML eng | $680K–$1.2M | $870K | +26% |
| GPU scheduling / cluster ops | ML eng | $620K–$1.05M | $780K | +9% |
| Data pipeline engineering | ML eng | $640K–$1.0M | $760K | +8% |
| Eval infrastructure | ML eng | $640K–$1.05M | $790K | +19% |
| Model serving / quantization | ML eng | $650K–$1.05M | $800K | +15% |
| AI SRE / reliability | ML eng | $620K–$990K | $740K | +21% |
| RAG & retrieval engineering | Applied | $420K–$780K | $560K | +11% |
| Agent product engineering | Applied | $460K–$820K | $610K | +24% |
| Evals & quality engineering | Applied | $400K–$740K | $540K | +18% |
| Prompt / structured outputs | Applied | $380K–$700K | $500K | +8% |
| AI product engineering | Applied | $400K–$720K | $520K | +9% |
| Fine-tuning & adaptation | Applied | $440K–$800K | $580K | +14% |
| Document AI & extraction | Applied | $420K–$740K | $550K | +12% |
| Voice / multimodal applied | Applied | $440K–$780K | $570K | +21% |
| Frontier-lab new-grad MTS | Early-career | $320K–$520K | $420K | +10% |
// Levels × tracks heatmap
Median total comp from intern to principal.
Median offers at five level bands across four engineering tracks, US-based. Cell intensity reflects relative comp. PhD candidates land at the upper end of each early-career band.
// By geography
Senior IC offers across 8 geographies.
Median total comp for senior IC roles in each geo, normalized to local currency, then USD-equivalent for reference. European markets have closed roughly 60% of the historical gap to US comp over the past 18 months.
Senior IC — cross-track median by geo
Cross-track because per-geo per-track samples thin out — see the full data in your account.
| Geography | Local currency | USD-equiv | vs. US median | YoY |
|---|---|---|---|---|
| United States – SF Bay | $760K–$1.2M | $920K | 100% | +12% |
| United States – NYC / Seattle | $680K–$1.05M | $820K | 89% | +11% |
| United Kingdom – London | £500K–£780K | $680K | 74% | +18% |
| France – Paris | €460K–€720K | $620K | 67% | +22% |
| Canada – Toronto | CA$720K–CA$1.05M | $640K | 70% | +13% |
| Singapore | S$640K–S$980K | $580K | 63% | +15% |
| India – Bengaluru | ₹2.1Cr–₹3.8Cr | $310K | 34% | +34% |
| UAE / Saudi (MENA) | AED 1.8M–2.6M | $540K | 59% | +28% |
// What moved — 12 months
Eight trend signals from the dataset.
The numbers worth knowing if you're planning a move in the next two quarters. Sourced from the May 2026 dataset against the year-ago window.
Median comp for inference engineers -- the role that ate the cost-center problem.
Applied AI's fastest-rising specialization. Agent product engineering.
Median comp for alignment & safety research. The bar moved fast as frontier labs scaled their safety orgs.
YoY median for India-based senior IC offers. Sarvam, Krutrim, and frontier-lab remote roles pulled the median up sharply.
Of senior IC offers in the dataset included a signing bonus over $200K -- up from 21% in 2025.
Median for prompt & structured outputs engineers. The Applied AI specialization with the slowest comp growth.
Increase in cross-lab senior engineering moves vs. the same window last year.
Median frontier-lab senior-IC offer cycle -- scope to written offer. Was 27 days in 2024.
// What “total comp” actually means
The four-piece composition of a senior IC offer.
“Total compensation” hides important variation. Most senior IC offers we see break down across four buckets — and the mix matters more than the headline number.
Base salaryis the smallest part of senior IC frontier-lab offers — usually 25–35% of total. Equity and signing bonuses do the heavy lifting. The labs that pay highest-base tend to pay lowest-equity, and vice versa.
Equityat frontier labs is usually the largest single piece — 40–55% of total — and is paid as RSU-equivalents on a 4-year vesting schedule. At pre-IPO frontier labs (most of the ones we track), equity is the most volatile and the most consequential component to negotiate.
Performance bonusranges 10–20% of total and is targeted at 100% of base or lower. Some labs pay this as a quarterly bonus; others bundle it into equity refreshes.
Signing bonusappeared in 38% of senior IC offers in the May 2026 dataset, up from 21% the prior year. Where it appears, it's significant — median $250K, with notice-period buyout terms layered on top in roughly half of cases.
// SENIOR IC OFFER — MEDIAN COMPOSITION
// Negotiation insights
What network members typically negotiate for.
Across 4,800 senior IC offers in the May 2026 dataset, six negotiation patterns show up over and over. Knowing the patterns is half the negotiation.
// PATTERN 01
Equity refresh tier
Senior IC offers come with an implicit equity refresh schedule. Negotiate which tier you're calibrated to -- moving up one tier is worth more than 10% on base.
// PATTERN 02
Signing bonus & notice buyout
If your current notice is long, ask the lab to cover it. Frontier labs are increasingly willing -- 47% of offers in our dataset that asked for this got it.
// PATTERN 03
Start-date flexibility
Offers issued in a hot quarter come with implicit deadline pressure. Pushing for a 60–90 day start window has a near-zero cost to the lab and a real cost to you to give up.
// PATTERN 04
Level calibration
“Are you sure I’m at the right level for this scope?” — the most under-asked question. Strong candidates often have a real case for a level bump that doesn’t get tested unless they raise it.
// PATTERN 05
Team / scope
Comp isn’t the only lever. Negotiating the specific team, project, or scope on day one is often the highest-value win — and the easiest to give to the candidate.
// PATTERN 06
Documented comp bands
Bring the data. Members who cited a specific network-verified comp range during negotiation got 9% higher final offers, on average, than those who didn’t.
Every other comp dataset I'd seen was self-reported and three months out of date. OpenTalent's data was the first one I trusted enough to bring to a negotiation. I cited a specific median, asked for the equity tier above what they offered, and walked away with $140K more.
Senior post-training engineer — negotiated a frontier-lab offer Q1 2026
// Methodology
How the data is built.
Every number on this page is sourced from offers we've seen — directly, from network members. No self-reporting from public boards. No scraped LinkedIn data. The full methodology is below.
// Sourcing
- Network-verified offers only. Members of the OpenTalent network share their offers with us — privately, voluntarily, and under contractual confidentiality.
- No scraped data. We do not pull from levels.fyi, LinkedIn, or any other public source. Comp data on public boards is too noisy to be useful for senior IC negotiation.
- Two-party verification. Where possible, we verify the offer against the lab's hiring side. ~78% of senior IC offers in our dataset are two-party verified.
- Trailing 12 months. The May 2026 dataset uses offers from May 2025 to April 2026. Older data is archived but not used for current medians.
// Construction
- Total comp definition. Base + equity (RSU-equivalent at offer-time valuation) + target bonus + signing-bonus annualized over 4 years.
- Level normalization. Senior IC = 5–8 years experience or its frontier-lab equivalent (e.g., L5 at Anthropic, MTS at OpenAI, Engineer III at DeepMind).
- Geography conversion. Local currency at offer date; USD equivalent at the same day's FX rate. We don't smooth.
- Privacy & aggregation. No median is reported with fewer than 12 underlying offers. We round all reported values to the nearest $10K for senior IC offers, $5K for early-career, and $0.5K/mo for interns.
// Related
Where this data lives next door.
Three other places the network publishes comp and market signal — each with its own angle.
Get full data access. Free for network members.
Lab-by-lab medians, geo cuts, level breakdowns, and the full 5,800-offer dataset behind the page above.