From software engineer
Backend, full-stack or platform engineers with production experience who want to work directly with customers on AI deployments.
What already counts
- Shipping and running production code, with reviews, tests and releases
- Debugging real systems through logs, metrics and traces
- APIs, databases and the system design you already practise
- Working in a team on a large shared codebase
What to add
- Discovery: asking customers questions before proposing a design
- Applied AI with evals, guardrails and staged rollout
- Enterprise constraints: identity, security reviews, personal data, legacy systems
- Writing for executives and non-engineers, and saying no well
- Comfort with work that has no ticket and no product manager
1See the job clearlyWeeks 1 to 2
Understand how FDE work differs from product engineering, and score yourself against real postings so the rest of the plan targets your gaps.
Show for it: A filled-in readiness checklist with a score and one line of proof per item, and a six-week plan aimed at your three biggest gaps.
2Applied AI you can proveWeeks 3 to 6
Build retrieval and an agent on realistic data, measure them with evals, and know their cost and latency at real volume.
Learn
Is this a good AI use case?Answers over a customer's private dataAgents and tools in customer workflowsBuilding MCP servers for a customer's toolsEvals: golden sets, graders and regression suitesGuardrails and human reviewCost and latency budgetsPractise
Take-home: Product search that understands how people typeTake-home: A CRM agent with tools and an approval stepPrompt Lab: Design an evalCode Lab: Monthly cost estimator with a budget alertCode Lab: Permission-aware retrieval for Kaveri BankCode Lab: A tool router for Raftaar's delay desk agentShow for it: Two finished take-homes, each with an eval report, a cost estimate at the customer's volume and a recorded walkthrough.
3Discovery and the customer roomWeeks 7 to 9
Learn to slow down at the start: find the real problem, map the people and scope a first project before writing code.
Learn
Running a discovery callFinding the real problem behind the requestDecomposing a vague problemWriting a scoping documentMapping the people: users, buyers, IT, security and legalSaying no and pushing backExplaining technical work to executivesPractise
Customer Room: A policy assistant the compliance team will acceptCustomer Room: "Use AI to reduce delays"Field Lab: Question list for a first call at Kaveri BankField Lab: Stakeholder map for Kaveri Bank's complaints projectField Lab: Two minutes with Kaveri Bank's CEO about a delayField Lab: Refuse Dukaan's auto-discount request with an alternativeShow for it: Three Customer Room sessions with a scorecard of solid or better on discovery and constraints, and one scoping document you would be happy to send a real customer.
4Enterprise systemsWeeks 10 to 11
Fill in the enterprise knowledge product teams rarely need: identity, private deployment, security reviews and personal data.
Show for it: Written answers to a five-question security review and a one-page data-flow note for a privacy officer, both reviewed by someone you trust.
5Tell the story and interviewWeek 12 onwards
Rewrite your experience as customer impact and practise the rounds product engineers usually fail: decomposition, customer design and the behavioural deep dive.
Show for it: A CV where every role has a line about users or customers, five behavioural stories written down, and three mock design rounds done.