Software Dev Engineer II, SIFT

Amazon Inc.; Boston (MA); On-site; USD 175,000 - 250,000; Full time

Job summary

Amazon Inc. is seeking a backend-focused AI-driven engineer in Boston to scale core systems that power field inputs into structured, trustworthy records. You will drive end-to-end design and ensure production-grade reliability in a fast-growing environment.

You will work with generative AI and large language models, integrating data sources safely, expanding MCP integrations, and mentoring peers to lift the whole team’s capability.

Qualifications

• Hands-on experience with generative AI and large language models is a strong prerequisite.
• Familiarity with Amazon Bedrock will set you up to make an impact fast.

Responsibilities

• Own features end to end — from the user interface down to the backend services behind them.
• Build and scale the services that turn raw, free-form field input into structured, searchable records.
• Expand the tools and integrations exposed through the Model Context Protocol (MCP).
• Wire large language models into the product and raise the quality bar on what they produce.
• Build the natural-language experiences that let people ask for what they need and get answers grounded in real data.
• Write the design docs, run code reviews, debug production issues, and mentor engineers.

Skills

Backend engineering

Generative AI

Large language models

System design

Tools

Amazon Bedrock

Job description

Every day, the people closest to our customers and partners learn things that matter — in meetings, in deal rooms, in the field. For years most of that went nowhere useful: stuck in an inbox, a memory, a note no one opened again. We changed that. We built a product that captures those raw field inputs, automatically finds the patterns hiding across thousands of them, and uses AI to turn what used to be hours of manual report-writing into minutes — and it's now part of how a large, fast-growing community does its job every day.

That's the part that makes this interesting: it already works, at real scale, for a lot of people — and that raises the bar rather than lowering it. The hard problems now are the ones you only get to when something is genuinely in production and genuinely depended on. How do you keep quality high as usage climbs and the data gets messier? How do you make AI-generated output people trust instead of double-check? How do you extend a system people already rely on without breaking what they count on? This is applied AI in the critical path of real work — where \"mostly right\" isn't good enough and someone notices the moment it's off.

That's where you come in.

Here's the kind of thing you'd own and solve:

• Scale what's already working. Take services that are live and heavily used and make them faster, sturdier, and ready for the next order of magnitude of users and data.
• Grow what the product can connect to. Expand the set of tools and integrations the system exposes through the Model Context Protocol (MCP), so the product can plug into more data sources and let AI assistants do more on a user's behalf — safely and reliably.
• Raise the quality bar on AI output. Improve how the product turns messy, free-form input into trustworthy, structured, searchable records — with smarter summarization, tagging, and quality gating so strong output ships and weak output doesn't.
• Extend the product thoughtfully. Add new capabilities and data sources to a system people already depend on, without disrupting the workflows they rely on every day.
• Make the trends sharper. Improve how the product spots patterns across a huge body of individual stories, so people see what's emerging without digging for it.
• Let people talk to their data. Build and refine the natural-language experiences that let someone ask for what they need — and get an answer built from real data, and only the data they're allowed to see.
• Keep it healthy. Own the reliability and operations of what you build, dig into production issues when they happen, and raise the operational bar for the whole team.

What sets this apart: you're joining something that already succeeded — and the interesting work is everything that comes next. You get the best of both worlds: the real users, real feedback, and real impact of a product in heavy use, plus a wide-open set of hard problems around scaling it, deepening the AI, and taking it places it hasn't gone yet. You'll make architectural calls that matter, see your impact quickly because the people using it will tell you exactly what's working, and be expected to lift the engineers around you — mentoring, reviewing designs, and being someone who makes the whole team sharper, not just your own code.

We're looking for someone who's built and operated real backend systems at scale, is comfortable improving a live product without breaking it, and is genuinely excited about applying AI to a hard, valuable problem rather than chasing the buzz. Hands-on experience with generative AI and large language models is a strong prerequisite — and familiarity with Amazon Bedrock in particular will set you up to make an impact fast. If you like owning things end to end, caring about the person on the other side of the screen, and raising the bar on something people already love — you'll feel at home here.

Key job responsibilities

You'll own features end to end — from the user interface down to the backend services behind them — designing, building, and running them in production, where real people rely on them every day.

• Own whole features from the screen a user sees to the services and data underneath, making the design calls across the full stack and shipping each one to production.
• Build and scale the services that turn raw, free-form field input into structured, searchable, trustworthy records — keeping them fast and reliable as usage grows.
• Expand the tools and integrations the product exposes through the Model Context Protocol (MCP), connecting it to more data sources and letting AI assistants safely do more on a user's behalf.
• Wire large language models into the product and raise the quality bar on what they produce — sharper summarization, tagging, and quality gating so strong output ships and weak output doesn't.
• Build the natural-language experiences that let people ask for what they need and get answers grounded in real data — and only the data they're allowed to see.
• Write the design docs, run code reviews, debug the gnarly production issues, and mentor the engineers around you.

A day in the life

You're working where solid engineering meets applied AI — shipping real features people use the moment they go out.

You might push a feature you own from the screen a user sees down to the service behind it: write the code, get it reviewed, shepherd it through testing into production. A good chunk of your time goes into making the AI better — tuning how it drafts and summarizes, then checking the output against real examples to prove it improved. You'll join a short daily standup and review teammates' code.

Some days stretch further — debugging production issues, or prototyping an idea.

About the team

We build AI that changes how the sales field works. Our users are the people talking to AWS customers and partners every day — and our job is to turn the stories they gather into something they can use right away. We capture their input, surface the patterns across