The Visa, the Salary, and the Japanese You Actually Need

India Japan Kaizen — AI Bridge Series, Part 3

Part 1 of this series sized the gap: Japan's AI investment and 220,000-person IT shortfall against India's AI talent surplus and the highest net talent outflow in the world. Part 2 named the actual companies already hiring — Fujitsu, NTT Data, Graphcore, Sarvam and Preferred Networks — and the agreements wiring the two governments together. This part answers the question an individual engineer actually asks before any of that matters: what does the move cost, what does it pay, and how much Japanese do you really need? A visa-policy clarification this April, and a handful of published placement stories, give the clearest answers yet.

The Visa Math: 70 Points Can Buy Three Years to Permanent Residency Instead of Ten

Most AI and software engineers from India enter under Japan's Highly Skilled Professional (HSP) status, category (i)(b) for Advanced Specialized/Technical Activities, rather than the standard Engineer/Specialist in Humanities visa. It runs on a points system, and clearing 70 points unlocks permanent residency in three years instead of the standard ten; 80 points gets there in one.

• Education. A doctorate is worth 30 points, a master's 20, a bachelor's 10.
• Experience. 20 points for 10+ years in the field, down to 5 points for three to four.
• Salary. Up to 40 points, scaling from zero below ¥3 million a year to the full 40 at ¥10 million and above.
• Age. Up to 15 points, highest under 30.
• Bonus points. A Japanese university degree adds 10; JLPT N2 adds 10, N1 adds 15; a recognized IT certification adds another 5.

For a mid-career engineer with a master's, five years of experience, and a competitive Tokyo AI salary, 70 points is well within reach before counting a single Japanese-language bonus point — which is where the next piece of news actually matters.

Japan Just Clarified Who Needs Japanese — and It Isn't Engineers

On 15 April 2026, Japan's immigration authorities revised the screening guidelines for the Engineer/Specialist in Humanities/International Services visa category. Applicants whose work is mainly interpersonal, language-based work — interpretation, customer-facing sales, hospitality — now need to show CEFR B2-equivalent ability, roughly JLPT N2. Pure technical and engineering roles are explicitly carved out of that requirement: writing code, building models, and running infrastructure isn't classified as language-based work, so an Indian AI engineer doesn't need to prove N2 to get the visa itself. Companies in the government's Category 1 and 2 employer tiers — larger, listed firms, which is where Fujitsu, NTT Data, and Graphcore from Part 2 all sit — face lighter document checks than Category 3 and 4 employers under the same revision.

That's a visa-eligibility floor, not a hiring bar. Nothing in the revision stops an individual company from preferring conversational Japanese for team meetings, and the HSP points above still reward N2 and N1 directly. What changed is that the government's own immigration paperwork no longer treats writing software as language work — a narrower, more accurate read of what an AI engineer's job actually is.

The Pay Gap That Makes the Case on Its Own

However the visa and language questions resolve, the number that actually moves people is pay. Industry salary data puts Japan's average AI engineer compensation at roughly ¥6.4–9.9 million a year (about $64,000), against an Asia-wide average closer to $17,000–21,000 for the same role in India. That's a three-to-four-times gap in a single move, before counting Japan's HSP fast-track to permanent residency or the spouse work authorization that comes with it. It's also the plainest explanation for why Part 1's 220,000-person shortfall and India's talent outflow keep showing up on opposite sides of the same ledger.

What the Move Has Actually Looked Like for People Who've Made It

Placement programs that specialize in matching international candidates with Japanese employers have started publishing some of these journeys directly. One recent account from FAST OFFER International follows an Indian IT engineering graduate who had struggled to land a job at home with no prior experience — she started Japanese lessons, and within six months was already passing mock interviews for AI engineering roles, landing an offer from a company willing to hire a fresh graduate. Another profiled candidate left Japan years earlier and spent nearly two decades away before returning specifically for an AI-focused role at a company standing up a brand-new AI department — citing language ability built over that time, and a genuine, specific interest in the country, as what actually closed the gap between an application and an offer.

Neither story involves a company from Part 2's named list, which is itself telling: the pipeline bringing individual Indian engineers into Japanese AI roles today runs mostly through private placement and language-training programs, not yet through the Fujitsus and NTT Datas doing the highest-profile hiring.

What's Missing

Every piece now points the same direction. The points system rewards the profile a strong Indian AI engineer already has. The April 2026 clarification removes a language bar that technical roles were never supposed to clear in the first place. The pay gap alone justifies the move. And real people are already making it. What's still missing is a single channel connecting those three facts to a specific company and a specific visa category — the METI-IT ministry employment-information exchange that Part 2 covered has been signed, but nothing public yet tells an individual engineer in Pune how to use it. Until it does, the route most people are actually taking runs through private placement agencies charging their own fees, not the bilateral channel built to replace them.

If you're an AI engineer evaluating this move, or weighing HSP points against an offer in hand — get in touch. Matching the right person to the right visa category and employer is exactly the kind of introduction India Japan Kaizen wants to make.

This is Part 3 of India Japan Kaizen's AI Bridge Series, publishing weekly. Part 4 will look at the reverse direction — the Japanese engineers and companies building AI teams inside India itself, and what day-to-day collaboration inside Bengaluru's new research campuses actually looks like.