Software Architect

Job description

**Architect --- Search** Myntra \| Search \& Discovery \| Bangalore **ABOUT MYNTRA:** Myntra is India's leading fashion and lifestyle e-commerce platform, serving millions of customers daily. Our Search \& Discovery team powers the first and most critical touchpoint on the app --- connecting customers to the right products across a catalogue of 20M styles. We operate at the intersection of large-scale ML, real-time systems, and deep domain knowledge of fashion **.** **THE ROLE:** We're looking for an E5 Architect for Search Relevance --- someone who combines deep ML and NLP expertise with the systems thinking to stitch it all together. This is not a pure research role. You will own the end-to-end relevance stack: from query understanding and semantic retrieval through ranking models and experimentation infrastructure. You'll be the technical anchor for the team, setting the direction for how Myntra's search understands, interprets, and serves intent at scal e.You'll work closely with engineering, product, and data teams to translate ML breakthroughs into production systems that directly impact GMV, conversion, and customer experience **WHAT YOU'LL DO** **Search Intelligence \&** * NLP Design and own the query understanding pipeline: intent classification, category prediction, attribute extraction, query rewriting, and spell correction. * Build and fine-tune LLMs and task-specific models for fashion-domain query comprehension --- handling the full spectrum from head queries to long-tail and zero-result cases. * Own the autocomplete experience end-to-end: suggestion models, personalised typeahead, trending and session-aware completions, and quality evaluation. * Drive hybrid search --- defining the architecture for blending dense (bi-encoder, cross-encoder) and lexical signals, and owning the embedding models that power semantic retrieval. * Tackle low-recall and zero-result queries through query expansion, taxonomy mapping, and semantic fallback strategies * Drive rigorous evaluation --- offline judgment sets, NDCG/MRR baselines, and A/B experiments --- to ensure every model change ships with evidence, not just intuition. * Stay ahead of the GenAI curve --- rapidly evaluate emerging LLMs and retrieval techniques, prototype and validate them, and drive a fast experimentation cadence that reduces time-to-production for new search capabilities. * Automate evaluation, data labelling, and experimentation workflows to compress the cycle from research idea to production feature. **Systems \& Architecture** * Own the end-to-end technical architecture of the search relevance stack --- from retrieval and indexing through model serving and online inference. * Design and drive the model serving infrastructure: online feature computation, model versioning, A/B traffic routing, shadow deployments, and latency budgets. * Own the catalogue enrichment and signal generation pipelines that feed retrieval and NLP models --- attribute tagging, taxonomy classification, and feedback loops. * Make and defend build-vs-buy decisions across the stack, and set the engineering bar for how ML systems are built, tested, and deployed in production **WHAT WE'RE LOOKING FOR** **Must-have** * 10--12 years of industry experience, with at least 4 years building production search relevance or recommendation systems at scale * Strong ML fundamentals: ranking, retrieval, classification, embedding models, and neural networks. * Hands-on NLP experience: transformer-based models (BERT, T5, or similar), fine-tuning workflows, and evaluation methodology. * Conceptual familiarity with search engines (Solr, Elasticsearch, or OpenSearch) --- enough to reason about index design, retrieval trade-offs, and relevance signals in collaboration with platform teams. * Proficiency in Python for model development; ability to read and review production code in Java/Scala or Go. * Proven ability to work across system boundaries --- translating ML requirements into engineering specs and vice versa. * Experience with LLM-based query rewriting, RAG, or generative search patterns * Strong systems sensibility --- you can reason about distributed systems, data pipelines, and serving infrastructure, and translate ML requirements into concrete engineering space * Hands-on experience designing or operating model serving infrastructure: online feature stores, low-latency inference, and deployment pipelines * Strong experimentation discipline: you know the difference between statistical and practical significance, and you ship with evidence **Good to have** * Hands-on experience with search engines (Solr, Elasticsearch, or OpenSearch) --- schema design, query DSL, or relevance tuning. * Experience with vector databases (Qdrant, Milvus, Vespa) and ANN systems (FAISS, ScaN) * Familiarity with knowledge graphs or product taxonomy systems in e-commerce. * Publications or applied research contributions in IR, NLP, or RecSys **WHO YOU ARE** * You think in systems, not just models. You can draw the full architecture on a whiteboard and then go implement any layer of it. * You move fast with GenAI --- you prototype with new models, automate evaluation and workflow steps, and consistently cut the time between idea and production. * You are deeply curious about the fashion and lifestyle domain --- and you know that relevance is ultimately a human problem. * You hold a high bar for rigour: you don't ship a model without understanding why it works. * You communicate complex trade-offs clearly to engineers, PMs, and leadership --- and you influence without authority. * You are energised by ambiguity; you bring structure to hard, open-ended problems. **IMPACT** Search at Myntra handles hundreds of millions of sessions per month. Every improvement you build --- whether a better ranker, a smarter query parser, or a tighter feedback loop --- ships to tens of millions of customers. At E5, you are expected to define the technical roadmap for relevance, men