Asks for

The ad says“Proficiency in Python; systems-level fluency (C++ / Go / Rust) for performance-critical paths.”
The ad says“Experience with distributed training and the frameworks behind it — PyTorch FSDP, DeepSpeed, Megatron, or Ray.”
The ad says“Experience with big-data and streaming stacks — Spark, Flink, or similar.”
The ad says“Proficiency in Python; systems-level fluency (C++ / Go / Rust) for performance-critical paths.”
The ad says“Depth in inference optimization: GPU kernel tuning, quantization, speculative decoding, KV-cache and memory/IO optimization. CUDA / GPU programming experience is a strong plus.”
The ad says“Experience running GPU fleets in production — Kubernetes (ideally GKE), GPU scheduling and allocation, and multi-region/multi-cluster deployment.”
The ad says“Depth in inference optimization: GPU kernel tuning, quantization, speculative decoding, KV-cache and memory/IO optimization. CUDA / GPU programming experience is a strong plus.”
The ad says“Strong hands-on experience with the modern LLM inference stack — TensorRT-LLM, vLLM, SGLang — and with production, low-latency model serving at scale.”
The ad says“Familiarity with building LLM-powered agents and agentic workflows, and a point of view on where autonomy can replace manual engineering toil.”
The ad says“Experience with distributed training and the frameworks behind it — PyTorch FSDP, DeepSpeed, Megatron, or Ray.”
The ad says“Depth in inference optimization: GPU kernel tuning, quantization, speculative decoding, KV-cache and memory/IO optimization. CUDA / GPU programming experience is a strong plus.”
The ad says“Proficiency in Python; systems-level fluency (C++ / Go / Rust) for performance-critical paths.”
The ad says“Experience with distributed training and the frameworks behind it — PyTorch FSDP, DeepSpeed, Megatron, or Ray.”
The ad says“Depth in inference optimization: GPU kernel tuning, quantization, speculative decoding, KV-cache and memory/IO optimization. CUDA / GPU programming experience is a strong plus.”
The ad says“Experience with distributed training and the frameworks behind it — PyTorch FSDP, DeepSpeed, Megatron, or Ray.”
The ad says“Proficiency in Python; systems-level fluency (C++ / Go / Rust) for performance-critical paths.”
The ad says“Strong hands-on experience with the modern LLM inference stack — TensorRT-LLM, vLLM, SGLang — and with production, low-latency model serving at scale.”
The ad says“Experience with big-data and streaming stacks — Spark, Flink, or similar.”
The ad says“Depth in inference optimization: GPU kernel tuning, quantization, speculative decoding, KV-cache and memory/IO optimization. CUDA / GPU programming experience is a strong plus.”
The ad says“Strong hands-on experience with the modern LLM inference stack — TensorRT-LLM, vLLM, SGLang — and with production, low-latency model serving at scale.”
The ad says“Strong hands-on experience with the modern LLM inference stack — TensorRT-LLM, vLLM, SGLang — and with production, low-latency model serving at scale.”

About Meesho

Meesho is India's fastest-growing internet commerce company, on a mission to democratize e-commerce for everyone. We serve millions of customers and over 1.75 million sellers through technology-driven innovation, building the scalable systems that power Meesho's most critical surfaces — Search, Recommendations, Personalized Ranking, Logistics, Fraud Detection, and Image Match.

The AI Platform sits at the heart of this. It serves a peak of 1M+ real-time deep-learning model inferences per second on ordinary days, scaling 3x+ on sale days — with the reliability that scale demands. The team works at the frontier of applied AI and infrastructure — multi-region inference, novel embedding-search algorithms, and optimized open-weight LLM models — squeezing out every bit of computation and passing the cost savings straight back to customers.

About the Role

We are looking for an experienced Engineering Manager – AI Engineering to lead the development of scalable AI platforms and infrastructure while managing high-performing engineering teams. You will drive the design, delivery, and optimization of production-grade AI systems powering AI use cases across Meesho.

What You'll Do

Lead, mentor, and grow a team of AI engineers — setting technical direction, raising the engineering bar, and owning execution and delivery end to end.

Architect and scale Meesho's AI platform: cross-region model inference, multi-GPU fleet allocation and management, distributed training, and feature-engineering infrastructure.

Drive inference optimization across the full stack — GPU kernel tuning, quantization (including outlier/tail-distribution handling), and memory/IO-bandwidth optimization — while building agents that codify and delegate known optimization procedures.

Optimize open-weight models at both the model and inference-engine level — distillation, quantization, speculative decoding, KV-cache and serving-engine tuning.

Scale data-science productivity through autonomous, agent-driven workflows spanning feature engineering, model training, and rollout.

Push the frontier across MLOps, LLMOps, compute efficiency, and distributed ML systems.

Partner with Product, Data Science, and Platform teams to turn AI capabilities into production impact for millions of users.

Own the team's operating rhythm: hiring, performance management, sprint planning, and OKRs.

What You'll Need

Bachelor's or Master's in Computer Science or a related field.

9+ years of software engineering experience, including 2+ years managing engineers.

Strong hands-on experience with the modern LLM inference stack — TensorRT-LLM, vLLM, SGLang — and with production, low-latency model serving at scale.

Depth in inference optimization: GPU kernel tuning, quantization, speculative decoding, KV-cache and memory/IO optimization. CUDA / GPU programming experience is a strong plus.

Experience with distributed training and the frameworks behind it — PyTorch FSDP, DeepSpeed, Megatron, or Ray.

Experience running GPU fleets in production — Kubernetes (ideally GKE), GPU scheduling and allocation, and multi-region/multi-cluster deployment.

Familiarity with building LLM-powered agents and agentic workflows, and a point of view on where autonomy can replace manual engineering toil.

Experience with big-data and streaming stacks — Spark, Flink, or similar.

Proficiency in Python; systems-level fluency (C++ / Go / Rust) for performance-critical paths.

Strong leadership, problem-solving, and stakeholder-management skills.

Preferred

Open-source contributions to inference engines, training frameworks, or ML infra tooling.

Experience managing GPU cost/efficiency (FinOps) for a large fleet on Cloud and Neo-Clouds.

Track record building platforms for high-scale consumer products (millions of users).

Familiarity with observability and reliability for ML systems (SLOs, autoscaling, incident response).