As of November 2, 2026, Sarvam AI lists 46 open roles, with 37 in Bengaluru. The ramp spans infrastructure, engineering, models, and product, with immediate application windows. Core hiring targets performance engineering, large language model training and inference, and voice‑agent systems built for Indic speech. Apply only through the official careers portal to be considered.
The depth of roles signals serious systems work. Infrastructure needs DevOps, Site Reliability for high‑performance computing, and performance engineers for inference and kernels. Models seeks training‑infrastructure talent. Required skills repeatedly include CUDA, Triton, PTX, Nsight, SGLang, vLLM, TensorRT‑LLM, Kubernetes, AWS, Terraform, and distributed training frameworks like Megatron‑LM and DeepSpeed.
Sarvam AI hiring: which Bengaluru teams are scaling now
Today’s Bengaluru-heavy openings cluster in Infrastructure (five roles), Engineering (twelve roles), Models (three roles), and Product (five roles). Marketing adds seven Bengaluru seats, while most Sales and Business Development roles sit in Delhi. Inference postings explicitly reference Indic automatic speech recognition and text‑to‑speech, underscoring demand for voice systems expertise.
Skills in demand for LLM, infra and speech roles
Kernels engineers must author custom CUDA and occasionally hand‑write PTX, profile with Nsight, and tune attention kernels; GitHub contributions are weighted strongly—“For this role, the GitHub filter is the highest‑yield signal.” Inference engineers modify SGLang or vLLM internals and own TTFT, throughput, and p99. Kubernetes, Terraform, and AWS fluency remain table stakes.
What Sarvam AI interview loops test
Expect loops to probe end‑to‑end systems thinking, not surface familiarity. Inference roles measure outcomes like TTFT, throughput, GPU utilization, and cost per million tokens. A recent candidate account described building a voice activity detector under time pressure—evidence of hands‑on, production‑minded evaluation. Prepare to defend decisions with numbers and profiles.
How to apply to Sarvam AI careers quickly and safely
Apply via sarvam.ai/careers; applications route through Ashby. Submit a focused resume with keywords reflecting your target stack—CUDA, Triton, PTX, SGLang, vLLM, TensorRT‑LLM, Kubernetes, Terraform, DeepSpeed—and link to meaningful code or kernels. Follow up within 48 hours with a concise portfolio note. Scam alert: avoid third‑party forms, fees, or unofficial messages; verify @sarvam.ai or Ashby links.
