Nykaa Interview Questions 2026

Tip: Nykaa: beauty and fashion e-commerce. Engineering focus: product catalog and search (100K+ SKUs), personalisation (shade finder, skin type recommender), supply chain (multi-warehouse inventory), D2C brand platform, and Nykaa Fashion (apparel). Interviews lean toward full-stack and data engineering roles.

Tip: Input: skin tone (from camera/picker), undertone (warm/cool/neutral), lip condition. Feature engineering: embed products by shade hex code + ingredients. Model: KNN on product embeddings to find nearest shades for skin tone. Cold start: map skin tone to shade families using a color wheel algorithm. A/B test: shade recommendations vs generic top-rated.

Tip: Inventory: per-SKU, per-warehouse stock levels. On order: reserve inventory in the nearest warehouse to the delivery address (reduces shipping cost/time). On fulfillment: deduct actual stock. Oversell protection: atomic DECREMENT-IF-POSITIVE. Reorder point triggers: stock < safety_stock → create purchase order. Sync with supplier via EDI.

Tip: Use ROW_NUMBER() OVER (PARTITION BY category_id ORDER BY units_sold DESC). Filter WHERE rn <= 5. Nykaa analytics rounds are SQL-heavy — window functions (ROW_NUMBER, RANK, DENSE_RANK, NTILE) appear in almost every interview. Practice on LeetCode SQL problems.

Tip: "Complete the look" = complementary product recommendation. Approach: (1) Association rules: products frequently bought together (Apriori algorithm). (2) Editorial: curated looks by beauty editors tagged in a CMS. (3) Hybrid: ML model scoring editorial looks by user affinity. Show both the ML and editorial dimensions.

Tip: Hypothesis: "showing shade swatches above the fold increases add-to-cart rate." Split: 50% control (current layout) / 50% treatment (new layout). Metrics: primary = add-to-cart rate; guardrails = bounce rate, checkout completion. Duration: minimum detectable effect at 80% power → calculate sample size. Avoid: multiple comparisons without correction.

Tip: Nykaa has strong mobile traffic — images are heavy for beauty products. Show you optimised: WebP format, lazy loading, responsive images (`srcset`), skeleton screens for perceived performance, minimising JavaScript bundle. Mention specific metric before/after: 'reduced LCP from 4.2s to 1.8s.'

Tip: Preparation: load test 2 weeks prior, pre-warm caches, autoscaling groups at minimum 5× baseline. During sale: virtual queue for checkout (Nykaa uses a waitlist to protect the checkout funnel). Graceful degradation: disable personalisation, serve cached product pages, reduce image quality. Post-event: review SLO compliance, identify bottlenecks.

Tip: CDN: globally distributed edge servers cache static assets (images, CSS, JS) close to users. Benefits: lower latency (nearest edge node), reduced origin load, higher availability. For Nykaa: 100K+ product images, each with 4 variants (thumbnail, medium, large, zoom). Without CDN, every image request hits origin in Mumbai — slow for users in Delhi/Chennai.

Tip: Approaches: (1) Edit distance (Levenshtein) at query time — expensive at scale. (2) Elasticsearch fuzzy query (fuzziness: 'AUTO') — BK-tree internally. (3) n-gram index: split terms into character n-grams and index separately. Nykaa's catalog has brand names like 'Kérastase' and 'L'Oréal' — Unicode normalisation is important.