Paste a 17-character VIN. The engine parses the WMI (positions 1-3), the manufacturer model code (positions 4-6), and the model-year letter (position 10), then resolves the car against a battery / warranty / MIDC-range table where each value carries its own citation. The default VIN below is the one from your brief, MAT629228MKC22274.
Four deterministic steps plus a sourced lookup. Nothing here is hard-coded to a single VIN, you can change any character and watch the output move.
A VIN is exactly 17 characters and never contains I, O, or Q. Both rules are enforced before anything is read, so a malformed VIN fails loudly instead of returning a confident wrong answer.
MAT resolves to Tata Motors, MA1 Mahindra, MAL Hyundai, MZ7 MG, and so on. India codes sit in the BIS-assigned MA block; each is tagged verified, reference, or research by source strength.
Positions 4-6 are the manufacturer model code (629 = Tigor EV, per your Tata reference). Position 10 is the ISO model-year letter, computed live: M = 2021, N = 2022, P = 2023, and on through the cycle.
Model + year narrow the car to a row carrying gross/usable battery, MIDC range, and battery warranty (years and km). Every value links to where it came from, and where a pack can't be told apart from the VIN alone, the engine says so instead of guessing.
This slice proves the engine and the sourcing method on real data. Tata is built out to 9 rows; Mahindra, MG, Hyundai and a few others are seeded as proof of breadth. The full deliverable is all 16 manufacturers with a complete, cited spec table, that is the work itself.
One honest finding worth flagging up front: there is no free public API that structurally decodes Indian VINs. NHTSA's vPIC returns an error on MAT629228MKC22274 (manufacturer not US-registered), and VAHAN is registration-gated and masked. The model-code maps and the spec table are hand-curated per manufacturer, which is exactly the part that needs doing.