Forward Deployed Engineer (AI)
Also posted as: FDE · Forward Deployed AI Engineer · Deployment Strategist · Customer Engineer (AI) · Applied AI Solutions Engineer
A Forward Deployed Engineer sits inside the customer's problem: you fly out (or dial in) to a bank, a media company or a merchant, work out what they actually need, then build and ship the AI system in their environment — their data, their legacy integrations, their compliance rules — until it hits an agreed number in production. In India the hiring is led by AI-first startups (Sarvam AI, Cartesia, DevRev, realfast, Deployment Inc), platform companies with big India teams (Databricks, Razorpay, Handshake, Level AI) and now IT-services firms (Accenture, Cognizant, Turing), with 16 of 23 job ads in Bengaluru. The role exists because enterprises can buy a model but cannot deploy one, and it pays well above standard backend work — but 19 of 23 job ads want 3+ years of shipped production software, so it suits working engineers moving into AI, not freshers.
Capabilities employers ask for
How often employers ask for each capability, measured across the job descriptions behind this page. Click one to see what "knowing it" means, how to learn it, and how to prove it.
Communicate AI trade-offs to stakeholders
18 of 23 job ads make communication an explicit requirement, not a soft bonus — Deployment Inc wants 'communication that holds up in a boardroom as well as a code review', Sarvam's dubbing FDSE role needs CXO/VP-level engagement, and Supervity calls polished communication 'mandatory'. This is the single most common line in FDE JDs.
Run a customer discovery → POC → pilot loop
17 of 23 job ads describe owning a customer engagement end to end — Sarvam's Deployment Strategist says 'from proof of concept through to the system running in production and hitting its committed KPIs', DevRev splits the job 30% customer-facing / 70% hands-on, and Razorpay makes you 'the single engineering face the merchant sees'. Discovery, POC, pilot, go-live is the job shape.
Write production-quality Python for AI work
14 of 23 job ads name Python, and Sarvam's senior FDSE role marks 'read, debug, and contribute to production FastAPI services and ML pipelines' as non-negotiable. TypeScript is the common second language (Handshake, DevRev, Deployment Inc) — but you cannot be a 'technical advisor' here, you write shipping code in front of the customer.
Build and consume REST APIs
13 of 23 job ads are really integration jobs: Razorpay wants 'APIs, webhooks, integration patterns'; DevRev adds GraphQL, pub/sub and large-scale data synchronisation; Ciroos wants REST APIs, webhooks and event-driven workflows against ServiceNow. Most FDE days are spent joining an AI system to somebody else's messy stack.
Deploy an AI service to the cloud
13 of 23 job ads ask you to run the system on AWS, GCP or Azure — and FDE deployment is harder than the usual kind: Cartesia wants 'cloud, VPC, and on-prem environments while navigating security, networking, and compliance', and Sarvam values BFSI / public-sector and on-prem experience. Learn one cloud deeply plus Docker, and understand what a customer's security review will ask.
Build a multi-step agent workflow
12 of 23 job ads want agents built and shipped, naming LangGraph, LlamaIndex, CrewAI or Google ADK — Sarvam asks for 'multi-agent architectures, including memory management and orchestration' and 'working AI agents in production — not a demo'. Sarvam, DevRev and Accenture also name MCP or agent-to-agent protocols, so learn tool calling and MCP alongside the framework.
Integrate LLM APIs into an application
12 of 23 job ads expect you to know LLMs as production components — Sarvam's phrasing is 'deeply understand how LLMs and agents work in production: context engineering, memory, RAG, tool use, structured outputs', and Deployment Inc wants 'LLM-backed systems that people actually used'. Both hosted and open-source models come up, because customers dictate which one you may run.
Translate business requirements into an AI solution design
11 of 23 job ads ask you to turn a vague business problem into a design: Level AI wants you to 'translate requirements into execution plans', Supervity to 'map complex customer workflows, architect bespoke AI solutions', and Deployment Inc to 'read a workflow and find where the value leaks'. This is what separates an FDE from a backend engineer with the same stack.
Deliver technical demos and answer RFPs
10 of 23 job ads involve demoing, presales or teaching: Neural Concept wants you to 'deliver proofs-of-concept demonstrating how our technology creates value' and train customer teams, Databricks wants experience 'communicating and/or teaching technical concepts to non-technical and technical audiences', Razorpay wants you to explain a complex system to a non-technical merchant lead. Practise a 10-minute live demo that survives a hostile question.
Build a grounded RAG application with citations
10 of 23 job ads name RAG — Accenture wants 'RAG architectures, vector databases, embeddings, and semantic search', Cognizant wants RAG pipelines with vector DBs and tokenisation. In FDE work the hard part is the customer's documents (PDFs, contracts, tickets, call transcripts), so practise ingestion and citation quality, not the framework tutorial.
Design scalable AI-backed systems
8 of 23 job ads ask you to 'reason about architecture, reliability, scale, trade-offs, and failure modes across services, databases, queues, and messaging systems' (Sarvam's exact words, with Kafka/SQS named); Handshake's senior role owns 'technical architecture for customer deployments'. Expect a design round about queues, retries and idempotency, not only about prompts.
Build voice or vision LLM features
7 of 23 job ads are voice-first, which is an India-specific edge: Sarvam deploys voice, WhatsApp, ASR/TTS and dubbing pipelines in Indian languages, Cartesia wants real-time voice/telephony/low-latency experience, Level AI runs AI virtual agents over telephony. If you want the highest-paying India FDE roles, one working voice or telephony integration is worth more than another chat demo.
Build an LLM evaluation harness
Only 6 of 23 job ads say 'evals' explicitly, but they are the most competitive ones: Sarvam asks you to 'set up eval pipelines, define metrics that matter for a given use case' in both its FDE and FDSE roles, and Deployment Inc wants you to 'set evaluation bar for work: baselines, regression gates and monitoring'. Because an FDE commits to a customer KPI, an eval harness is how you prove the deployment worked.
Ship faster with AI coding assistants
4 of 23 job ads already require daily use of coding agents — realfast wants you to 'operate in cockpit mode. You steer. Your agents execute. Claude Code, Codex, and others write code, run tests, open PRs', and Razorpay expects you to 'lean on coding agents for routine PRs'. Small share today, but it is how FDE teams justify one engineer covering several customer deployments, so expect it to be asked about.
Families: Product, business & communication · Programming foundations · Cloud, deployment & production · Agents & workflows · LLM application development · Retrieval & knowledge systems · Evaluation, safety & observability
"AWS" on a JD is not "learn AWS"
The words employers write, translated into what they want you to be able to do for this role.
Skip, for now
- Kubernetes — Named in 4 of 23 job ads (Cartesia, Ciroos, Sarvam's dubbing role, Handshake as a nice-to-have) and always next to years of platform or SRE experience. Docker plus one cloud deploy covers the 13 job ads that ask for deployment; add K8s only when a customer's VPC forces it.
- Fine-tuning and training your own models — Appears in 3 of 23 job ads, all at ML-platform vendors (Databricks, Turing x2). FDE work is about making an existing model work inside a customer's constraints — the job ads that mention fine-tuning list it after RAG, agents and evals, never before.
- Classical ML / PyTorch depth (pandas, scikit-learn, deep learning) — 7 of 23 job ads want ML background, but they cluster at Databricks, Turing and Neural Concept (which asks for a Master's/PhD plus CAD/CAE simulation). The startup and platform FDE roles — Sarvam, Cartesia, Razorpay, DevRev, Handshake, realfast, Deployment Inc — ask for none of it. Skip unless you are targeting the ML-vendor track.
- React / front-end frameworks — Only the 2 Handshake job ads require TypeScript + React. Everywhere else the customer-facing surface is an API, a webhook or someone else's UI. Basic JavaScript for integrations is enough; a full front-end specialisation is not what gets you hired here.
- Spark and big-data platforms — Databricks lists Spark as preferred, and no other job ad mentions it. SQL (8 of 23) and the ability to move a customer's data through queues and APIs matter far more than distributed compute frameworks.
Deploy an agent into a fake customer's stack — with a KPI, an eval bar and a runbook
Invent a plausible Indian customer (a lender, a D2C brand, an OTT platform) and write a one-page discovery note: their workflow today, the number you will move, and their constraints (data cannot leave their VPC, Hindi + English users, an existing ticketing tool). Then build the thing: ingest their messy documents and past tickets into a grounded RAG layer, wrap it in a LangGraph or ADK agent that calls two real integrations (a CRM or ticketing sandbox such as Freshdesk/Zoho/ServiceNow developer accounts, plus a webhook back into Slack or WhatsApp), expose it as a FastAPI service, containerise it and deploy it to one cloud. Finish it the way an FDE does: an eval set with a baseline and a regression gate, a monitoring dashboard, a runbook for the customer's ops team, and a 10-minute recorded demo that opens with the number, not the architecture. This mirrors the Sarvam, DevRev, Level AI and Deployment Inc job ads almost line for line.
- A written discovery note names one measurable KPI (e.g. first-response time, ticket deflection rate) with a measured before-and-after, plus an honest account of what did not work
- The agent completes a real multi-step task against at least two external systems with structured outputs, and stops for human approval before any write action
- Eval harness of 40+ cases runs in CI with a regression gate; the README shows one change that moved retrieval or task-success score, and the deployment survives a bad-input test
- Deployed service runs from a container on AWS/GCP/Azure with secrets in a secret store, request tracing, and a documented answer to 'can this run inside our VPC?'
- A 10-minute demo video and a one-page runbook (how it is run, monitored, and handed over) that a non-technical stakeholder could follow
What the interviews look like
The rounds you'll actually face, in the order they usually come.
- 1
Screening
Recruiter or hiring manager checks years of production software shipped (most job ads want 3-6+), whether you have been customer-facing before, your language stack (Python and/or TypeScript), cloud exposure and travel/on-site willingness — DevRev asks for up to 30% travel, Level AI for EST hours. Lead with a deployment you owned end to end and the number it moved.
- 2
Technical / take-home
Hands-on coding — usually a live session or short take-home building an integration or a small agent/RAG service, plus debugging across a distributed path (API to queue to worker to model to storage, in Sarvam's words). Expect to be asked how you used AI coding tools and to defend the code they wrote.
- 3
Solution design on a customer scenario
You are given a messy customer situation (legacy system, imperfect data, compliance constraint) and asked to design the deployment: retrieval and agent architecture, integrations, failure modes, evals, monitoring, on-prem/VPC options, cost and latency. Sarvam, Accenture and Handshake all screen for architecture and trade-off reasoning here.
- 4
Customer / stakeholder simulation
A role-play or demo round: run discovery with a 'customer', present a POC, and handle scope pushback or an unhappy stakeholder. Panels look for the Razorpay test — can you explain a complex system to a non-technical merchant lead — and for how you say no to a bad request without losing the account.
- 5
Values / ownership
Founder, delivery lead or bar-raiser conversation on high agency and ambiguity: a time you shipped without a spec, how you handled a deployment that missed its KPI, and how you work with a distributed US/India team. Several job ads (Sarvam, realfast, Deployment Inc) name high agency or comfort without a map as an explicit requirement.
What people ask before choosing this role
Can a fresher get a Forward Deployed Engineer (AI) job in India?
Realistically, no — not straight away. Only 0 of the 23 job ads behind this page accept 0–2 years; most want people who have already shipped software or run projects. It is a strong second move rather than a first job.
What is the salary of a Forward Deployed Engineer (AI) in India?
Entry-level roles cluster around ₹18–28 LPA, rising to about ₹90 LPA with experience. Every band on this page is quoted from a named source with a link, and pay varies widely by city, company type and whether the employer is an IT-services firm, a global capability centre or a product startup.
How long does it take to become a Forward Deployed Engineer (AI)?
The six capabilities employers ask for most add up to roughly 110 focused hours — about 14 weeks at 8 hours a week, if you are starting from zero on all of them. Most people are not: the self-check on this page works out what you can skip, which is usually a large part of it.
What skills do you need for a Forward Deployed Engineer (AI) role?
Across the 23 job ads behind this page, the most-requested capabilities are Communicate AI trade-offs to stakeholders (78% of ads), Run a customer discovery → POC → pilot loop (74% of ads) and Write production-quality Python for AI work (61% of ads). Note these are capabilities, not tools — employers write tool names, but what they are buying is the ability to do the work.
Which cities in India have the most Forward Deployed Engineer openings?
Bengaluru (16), Delhi NCR (3), Remote (India) (3) and Mumbai (1) — counted across the 23 job ads behind this page. Remote-India roles are counted separately where the ad said so.
Is demand for Forward Deployed Engineer roles in India growing?
Demand is growing faster than almost any other AI title: TeamLease Digital reports Indian FDE demand up ~800% across the first three quarters of 2025, and Indeed's global FDE listings rose 729% YoY to April 2026, against Naukri JobSpeak's +33% YoY for AI/ML jobs overall in July 2026. The India pool is still small in absolute terms (low hundreds of open roles), concentrated in Bengaluru and at AI-first startups such as Sarvam AI, which publicly opened FDE, FDSE and Deployment Strategist hiring this year.
Do I need a degree or a paid certificate for this?
Nothing on this page requires a paid certificate, and none of the 23 job ads behind it asked for one by name. What they ask for is evidence you can do the work — a public repo, a deployed project, something a hiring manager can open. That is what the path on this page is built to produce.
Companies with this role open in India
A sample of employers we saw hiring for this role — IT services, global capability centres, product companies and startups.
What it pays
How much demand
- 80'Forward Deployed Engineer (FDE)' jobs in India (August 2026) — a small, high-signal poolglassdoor
- 243Glassdoor on a broader keyword match showed 243 'forward deployed engineer' jobs in Indiaglassdoor
- 7Naukri surfaced only 7 FDE listings by title (Accenture x2, Info Edge, Insurity, Datavruti, DevRev x2) — most India FDE roles are posted on company career pages, not job boardsnaukri
- 5,330Indeed (global) FDE listings grew from 643 in Apr 2025 to 5,330 in Apr 2026 (+729% YoY); the India pool is estimated at ~250-270 open roles at any timeother
Where this role is heading
- Demand is growing faster than almost any other AI title: TeamLease Digital reports Indian FDE demand up ~800% across the first three quarters of 2025, and Indeed's global FDE listings rose 729% YoY to April 2026, against Naukri JobSpeak's +33% YoY for AI/ML jobs overall in July 2026.
- The India pool is still small in absolute terms (low hundreds of open roles), concentrated in Bengaluru and at AI-first startups such as Sarvam AI, which publicly opened FDE, FDSE and Deployment Strategist hiring this year.
Capability percentages come from 23 job descriptions read in full on 24-08-2026. How we do this