Sarvam AI

sarvam.ai · 33 open roles in the last 6 months · hiring in 1 country

Loading…

Sarvam AI appears to be investing most heavily in turning its India-focused model stack into production products: high-scale Model APIs, enterprise voice agents on Samvaad, and strategic-sector deployments under Chanakya, while continuing to build its own training and inference infrastructure.

Inferred from 42 job postings from the last year, with recent ones weighted more · updated 16 Oct 2026. Not a statement by the company.

What changed recently

The last one to three months add a stronger productization push on top of the earlier infrastructure buildout. Recent roles cluster around operating tens-of-millions-per-day Model APIs, running millions of voice calls on Samvaad, and adding GTM, PM, and partner-enablement capacity so enterprise and government deployments can scale beyond bespoke delivery.

Scalingmedium confidence

Adding enterprise GTM and marketing for complex India-specific buying motions

1m 23m 56m 71y 7

The commercial side is being built for long-cycle enterprise adoption, not just self-serve developer growth. Product marketing, solutioning, API GTM, growth marketing, and marketing ops roles point to a two-track motion: enterprise sales enablement for large institutions and funnel optimization for developers and technical buyers.

Product marketing roles explicitly mention sales enablement, ROI benchmarks, demos, and case studies
Marketing ops role introduces CRM, automation, attribution, and handoff discipline
Growth roles focus on API signup, activation, retention, and weekly experiment cadence
Enterprise target sectors recur: financial services, healthcare, public sector

Scalinghigh confidence

Owning the full model stack: training infrastructure, GPU fleet, and inference performance

1m 13m 36m 61y 6

Underneath the products, Sarvam appears to be building significant in-house model infrastructure. Roles cover distributed training infrastructure for the next family of foundational models, self-service GPU platforms, multi-vendor GPU fleet reliability, inference runtime optimization, custom CUDA kernels, speculative decoding, KV cache transfer, and cost-per-token improvements.

Recent training-infra role says it will own infrastructure for the 'next family of foundational models'
Separate roles target both training and serving, not just one side of the stack
Performance postings specify CUDA, Triton, PTX/DSL kernels, speculative decoding, and KV cache transfer
Infrastructure roles mention thousands of GPUs and a large multi-vendor fleet

Scalinghigh confidence

Building sovereign AI offerings for defence and strategic institutions through Chanakya

1m 03m 26m 61y 9

Chanakya stands out as a distinct vertical, aimed at defence, intelligence, and national-security users in India. The mix of PM, GTM, engagement, deployment, frontend, MLOps, and embedded infrastructure/data roles points to both product development and field delivery in constrained settings such as air-gapped, on-prem, offline-first, and classified environments.

Recent PM and GTM roles in Delhi add product and commercial structure on top of earlier deployment hiring
Roles repeatedly mention air-gapped deployments, offline-first operation, and constrained or classified environments
Hiring spans product, GTM, deployment, MLOps, frontend, and client-facing engagement rather than one isolated function
Target users are unusually specific: defence users, intelligence, strategic sector, central ministries, policy bodies

Scalinghigh confidence

Scaling real-time Model APIs into a core enterprise and developer product

1m 13m 36m 41y 4

The clearest current push is the developer- and enterprise-facing Model API platform. Recent roles call out tens of millions of API calls per day, real-time streaming, speech and multimodal serving, low-latency invocation, and commercial plumbing like metering, billing, prepaid wallets, and rate limiting; the PM and GTM roles suggest this is moving beyond infrastructure into a product and go-to-market surface.

Two recent roles directly on Model APIs in the last 2 months: platform engineering and product management
Posting names concrete commercial features: metering, billing, prepaid wallets, pricing, migrations
Scale is explicit: 'tens of millions of API calls every day'
Older staff role mentions a Python-to-Go rewrite, while the newest role emphasizes reliability, latency, and streaming operations

Scalinghigh confidence

Hardening Samvaad for enterprise-grade voice agents and outbound campaigns

1m 23m 26m 31y 3

Sarvam looks to be doubling down on Samvaad as a large-scale conversational AI platform. Two staff-level engineering hires in the last month focus on telephony backend, workflow orchestration, observability, integration testing, agent/campaign/user management, and sub-second latency for millions of personalized voice calls, which points to scaling a mature product rather than exploring a new one.

Two staff engineers were posted within 10 days of each other for the same platform area
Postings cite 'millions of voice calls a day' and '50M+ minutes of conversations every month'
Both roles stress observability and integration testing, suggesting reliability bottlenecks at current scale
Customer set spans enterprises, digital natives, governments, and public institutions

Scalinghigh confidence

Scaling deployment and partner delivery so enterprise AI goes from pilots to production

1m 03m 36m 41y 4

A second layer of investment sits between core product and sales: deployment strategy, agent engineering, solution architecture, and partner enablement. That combination suggests Sarvam is trying to make implementations repeatable, help customers get from PoC to live use, and train outside partners to deliver the stack without heavy dependence on Sarvam engineers.

A 'Technical Partner Enablement Manager' role says partner certification is early enough to 'own its shape'
Deployment roles emphasize moving customers from pilots/PoCs into production and delivering committed KPIs
Agent Engineer role focuses on live operations, evals, instrumentation, and adapting proven patterns quickly
Solution Architect and deployment roles indicate a solutions-heavy sales motion for enterprise and public-sector buyers

New betlow confidenceNo roles in 3 months

Exploring edge and on-device enterprise AI as a newer product bet

1m 03m 06m 11y 1

Compared with the heavier API, voice, and Chanakya hiring, on-device AI shows up as a narrower but notable product thread. The role frames it as a 'platform-defining' enterprise product with MDM integration, offline reliability, data sovereignty, and fleet management, which suggests a newer bet aligned with regulated buyers' deployment needs.

Single dedicated PM role labels it a 'platform-defining role'
Requirements are specific to edge deployment: offline reliability, MDM integration, enterprise security and identity
Theme fits the broader sovereign-AI and constrained-environment pattern, but has limited direct hiring evidence so far

Product Manager On-Device & Edge AI

Where they are expanding

Delhi — Hiring in Delhi clusters around Chanakya and strategic-sector work: GTM, product, engagement, deployment, MLOps, and customer-facing roles. That looks like a deliberate presence near defence, government, and national-security buyers rather than generic expansion.
Chennai — Chennai appears alongside Bengaluru in performance engineering and AI infrastructure roles tied to GPU serving, kernels, and platform infrastructure. It looks like a growing hiring pocket for low-level systems and infra talent, though still narrower than the Delhi signal.

Questions to ask in a coffee chat

How much of the current engineering priority is on stabilizing the Model API platform versus adding new model capabilities like speech and multimodal endpoints?
On Samvaad, are the biggest challenges now telephony scale and observability, or are teams still changing the core product workflow a lot?
For Chanakya, how much of the work is becoming reusable product versus one-off deployment engineering for each strategic customer?
How are partner enablement and deployment teams changing the way Sarvam ships enterprise projects compared with six months ago?
What changed internally that led to recent staff-level hiring around voice systems and Model APIs at the same time?
For the training and inference infra teams, are they mostly supporting existing model families, or preparing for materially larger next-generation foundation models?

What this can't tell you

Almost all roles are in India, so the data shows functional emphasis more clearly than international expansion plans.
Job postings reveal where the company needs capability, but not whether this is new headcount, replacement hiring, or hiring ahead of signed customer demand.
Some themes, especially on-device AI, rest on a small number of postings, so they may be exploratory rather than major company-wide bets.
The postings are strongest on engineering, product, and GTM intent; they are weaker at showing which product line is winning most internal attention once teams are staffed.