Grab interview, decoded.

Five to six stages over roughly three to four weeks, per Grab's own account of its process. It is a hard funnel: the company puts the share of candidates who start the loop and reach an offer at three to five per cent.

The first filter is a timed test on HackerRank or Codility: two to three problems, medium to hard, in 60 to 90 minutes, scored on passing test cases. Candidates who went through it describe it as noise removal rather than signal, and the problems as standard data structures and algorithms with a Grab flavour: delivery routes, ETAs, drivers near a pickup, transactions that time out. The exact set rotates. What follows are the problem classes candidates report, and the approach that solves each.

• Shortest route with traffic: edge costs that change with the time of day: Dijkstra where the cost of an edge depends on the time you reach it. Keep the priority queue on arrival time, and when you relax an edge, look up its cost at the current arrival time rather than a fixed weight. If waiting is allowed the answer is still monotone, so plain Dijkstra remains correct.
• Drivers nearest a pickup point (geospatial queries): Bucket drivers into a grid or geohash cells, search the cell and its neighbours, and keep the k closest in a max-heap of size k. Haversine only for the candidates that survive the bucket filter; sorting every driver by distance is the answer that times out.
• Calculating ETAs across a set of deliveries: Usually interval or greedy scheduling: sort by deadline or by start, sweep once, keep a heap of in-flight jobs. Read the input format twice; the trap in these is units (seconds versus minutes) and an inclusive versus exclusive end time.
• Transactions with timeouts and consistency requirements: Simulate a ledger with an idempotency key per request: a retry with the same key must not apply twice, a timed-out request must be reconciled rather than assumed failed. A hash map from key to outcome, and a queue of pending requests ordered by expiry, covers what the tests check.
• LRU cache: A hash map from key to node plus a doubly linked list; move a node to the head on access, evict from the tail. O(1) for get and put. Named in Grab's coding rounds; a clean version takes ten minutes, so write it from memory before the test.
• Trapping rain water: Two pointers from both ends holding the running maximum on each side; add the difference between that maximum and the current height on the lower side. O(n), no extra arrays. Also named in Grab's rounds.

Score comes from test cases passed, not from how you got there, so finish a working version of each problem before polishing any of them, and test the boundary the statement hints at (empty input, one element, the largest allowed n). A partial that passes half the cases beats an elegant solution that never ran.

Grab calls these the 4Hs and scores the behavioural round against them by name. The wording below is Grab's own, and it is worth reading closely: "Honour" is about ownership and following through rather than politeness, and "Humility" is explicitly about owning mistakes.

• Heart — being fiercely user-first, and doing what's right for users even when it's hard
• Hunger — moving fast and thinking boldly, testing ideas quickly and pivoting decisively
• Honour — radical ownership: challenge directly, debate hard, commit fully, follow through without exception
• Humility — owning mistakes and learning from them fast, and coaching others to raise the standard
• Working with people the team wants to work with, which Grab's own hiring write-up treats as a blocker rather than a tiebreak

The behavioural half maps onto the 4Hs, so bring stories that carry ownership and recovery rather than only wins. The technical half is where Grab differs most from a generic big-tech loop: the design questions are about its actual operating conditions.

Levels.fyi medians for Singapore, in Singapore dollars, read on 15 August 2026. The median across all levels sits around S$174K. Equity is RSUs on a four-year vest, and its share of the package climbs steeply at G5 and above — at G6 the stock is worth more than half the cash.

1. Prepare the ride-matching design properly: geospatial indexing, ranking, and what your system does with a location update that arrives ninety seconds late. Designing it for perfect connectivity is the most common way to lose this round.
2. Have real numbers for the currencies and countries you have worked with. Grab operates across eight markets and multi-currency handling comes up as a design constraint rather than a detail.
3. Practise coding while explaining, and be ready to hand-trace your loop. At least one reported round required tracing each input before running anything.
4. Bring one story per H, and make sure Honour and Humility are carried by a failure you owned rather than a success you narrate modestly.

This guide covers Grab's engineering and research hiring. For management and leadership roles the loop is similar but the bar shifts to people, delivery and strategy, so pair it with the leadership interview prep hub. The bar for your exact role comes from the role-by-role guides, and the prep that actually transfers is spoken, so run a mock interview before the real one.

Also interviewing at Shopee or TikTok? Their guides follow the same format, from the first screen to the offer.

• 01Grab, Form 6-K: majority stake in Atome Financialthe 60% stake for $1.49B in cash, the option on the remaining 40%, and completion expected by Q3 2027; announced 15 September 2026
• 02Grab Engineering, How Grab hires engineers in Singaporeofficial: the stages, the portfolio screen, the three-to-five per cent offer rate, and the culture veto.
• 03Grab Careers, Our valuesthe 4Hs in Grab's own wording, quoted above.
• 04Levels.fyi, Grab Software Engineer, Singaporethe G3–G6 bands and medians, read 15 August 2026.
• 05Glassdoor, Grab interview questions, Singaporecandidate-reported rounds and difficulty.
• 06Taro, Grab Senior Software Engineer, Singaporefirst-hand report, including the hand-tracing requirement and the HackerRank filter.
• 07TechPrep, Grab's interview processthe assessment on HackerRank or Codility, its domain-flavoured problems (delivery routes, transaction timeouts), and LRU Cache and Trapping Rain Water named in the coding rounds.
• 08OphyAI, Grab interview guidethe 60-to-90-minute window and the problem classes in the assessment section: time-varying shortest paths, geospatial queries, transactions with consistency requirements.

Interview processes change. This reflects widely-reported and sourced conditions as of 2026 — confirm specifics with your recruiter, and treat it as a map rather than a guarantee.