I build AI products developers rely on.
- From
- Nirbhay Singh
- Role
- Product Manager – Technical, AWS
- Before
- Senior TPM, AWS · Deloitte Consulting
- Focus
- AI agents · search · personalization
I spent nearly five years as a technical program manager before moving to product, so I'm at home in design reviews and experiment readouts.
~15K
Customer reports in scope
for the AI documentation agent
5.77M
Searches a year
run through personalized ranking
5×
More dark-mode usage
after the theme launch
$3M+
Saved each year
in technical writer time
01 / Work
Recent work, written up as memos.
01AI agents
An AI agent that fixes documentation errors
- Problem
- Customers report about 1,100 documentation errors a month. The open backlog had grown to about 15,000, far more than writers could work through.
- Bet
- An agent that reads each report, writes the fix, and sends it to the page owner to approve. A report only counts as resolved once the fix is live.
- Result
- Piloted on one service team's full backlog in August 2026. Owners in the pilot review fixes instead of writing them.
~15Kbacklog in scope
Read the memo 02Personalization
Personalized search for 5.77M queries a year
- Problem
- 49% of documentation searches failed, and 57% of people had to search more than once.
- Bet
- Re-rank results by role and experience level with Amazon Personalize. No sign-in required, and a 250 ms latency limit.
- Result
- Click-through went from 42.95% to 43.75% in an A/B test. The lift was statistically significant.
5.77Msearches a year
Read the memo 03Product strategy
One preview plan to replace four prototypes
- Problem
- In a UX audit, 70% of writers said previewing their work was a problem. Four teams were building separate prototypes to fix it.
- Bet
- Merge the four into one plan: previews in code review, generated by the existing analysis pipeline, with links writers can share.
- Result
- The plan set the roadmap for the second half of 2026, and the duplicate prototype work stopped.
4 → 1prototypes merged
Read the memo 04Generative AI
AI summaries in documentation search
- Problem
- People often opened several pages to answer one question, and leadership had committed to making the docs easier to use that year.
- Bet
- Show an AI-generated summary in search results, and roll it out in phases so we could watch quality as exposure grew.
- Result
- Shipped on the committed date. Later AI summarization work on the site built on it.
On timecommitted launch date
Read the memo 02 / About
I came to product from program management.
I'm a technical product manager at Amazon Web Services. I work on AWS Documentation, which millions of developers read every year. Right now that means an AI agent that drafts fixes for documentation errors. Before that I worked on personalized search and AI summaries in search results.
I moved into product after nearly five years as a technical program manager at AWS, mostly running migrations and launches. That work taught me to read a design doc closely and to ask what a feature will cost before it goes on the roadmap. Before AWS I spent four years at Deloitte Consulting, much of it on AI and data strategy for pharma and financial services clients.
I like to write the problem down first, agree on one metric that tells us it's fixed, and test small before rolling out. I bring legal, privacy, and security in early, which is usually what keeps a launch date intact.
- Product strategyI turn a vague customer problem into a written vision and a requirements doc engineering can build from.
- AI productsI've shipped an AI agent and generative search features, with a person reviewing output wherever a mistake would reach customers.
- Experiments and metricsI pick the metric before we build, then set up the A/B test and the dashboard to read it.
- Personalization and identityRanking by role, optional sign-in, and profile data models, with privacy review from the first draft.
- Technical depthNearly five years as a TPM on serverless migrations. I'm comfortable in a design review.
- LaunchesI've taken launches through legal, privacy, security, and app security review and kept the date.
03 / Experience
Nine years at AWS and Deloitte.
Apr 2026 – Present
Amazon Web ServicesNew York City
Product Manager – Technical
I own AI and platform products for AWS Documentation: the tools writers use and the AI that keeps the docs accurate.
- Product owner for an AI agent that triages customer reports, writes the documentation fix, and sends it to the page owner for approval. It's built for about 1,100 new reports a month and a backlog of about 15,000.
- Picked the agent's main metric (the share of reports it fully resolves) and changed the definition of done from merged to live on the site.
- Wrote the product vision for document preview, which combined four competing prototypes into one plan. A UX audit had shown 70% of writers struggled with preview.
- Took the authoring platform's new home page to general availability on its deadline after clearing problems with analytics, the AI assistant, and its safety filters.
- Built adoption and quality dashboards that leadership and partner teams use in place of manual reports.
Close+
Oct 2023 – Mar 2026
Amazon Web ServicesNew York City
Senior Technical Program Manager
Led the roadmap for AI features and customer-facing launches on docs.aws.amazon.com, which covers more than 200 AWS services. Most of this work was product work under a TPM title: finding the customer problem, deciding what to build, and measuring whether it worked.
- Search: 49% of searches failed across 5.77M a year. I wrote the requirements for role-aware re-ranking that works without sign-in and set a 250 ms latency limit. Click-through rose in a controlled A/B test with a statistically significant lift, and clicks spread beyond the top result.
- AI summaries: I scoped AI-generated summaries in search results in a decision doc and chose a phased rollout to limit the risk of wrong answers on a customer-facing site. It shipped on the date leadership had committed to.
- Sign-in: I chose AWS Builder ID over AWS account sign-in so students and people evaluating AWS could sign in without an account. I wrote the requirements and the profile data model, and brought identity, legal, privacy, security, design, and app security onto one plan.
- Site redesign: I shipped new site navigation and a visual refresh. Releasing the refresh across every content type at once, instead of in phases, put it in front of customers about 3 months early.
- Customer feedback: I prioritized small fixes developers had asked for, like a collapsible side panel and a broken YAML/JSON filter, which they then praised publicly on Reddit. I timed a device-matched dark theme to when engineering had capacity, and dark-mode usage went up 5×.
- New customers: I moved in-console tutorials to a faster system so customers in their first 90 days get help in under 500 ms, and wrote the PRFAQ for moving the Help Panel to serverless. API performance improved 40% and operational overhead dropped 50%.
- Platform calls: I wrote the decision doc and migration plan for a new docs authoring format, and launched the docs site in isolated government regions so restricted customers could use it.
Expand+
Jun 2021 – Oct 2023
Amazon Web ServicesNew York City
Technical Program Manager II
Ran the publishing platform behind docs.aws.amazon.com. My customers were the AWS technical writers who publish to it and the developers waiting on their updates.
- Writers were waiting too long for changes to go live. I led 5 engineers in moving deployment and hosting for docs.aws.amazon.com and docs.amazonaws.cn to Native AWS. Publishing got 75% faster, which saves more than $3M a year in writer time.
- I designed a serverless publishing flow on S3 and Lambda that made the pipeline 33% more efficient.
- I owned documentation launch readiness for re:Invent and other AWS conferences, so docs for new services went live with the announcements.
- I owned operations for the deployment and hosting infrastructure behind documentation for 200+ AWS services.
Expand+
Aug 2019 – Jun 2021
Deloitte ConsultingNew York City
Technology Consultant, then Senior Consultant
Program delivery for Fortune 500 clients in pharma and financial services.
- [Pharma] Led development of an AI platform that triages product complaints automatically. Processing time dropped 70% and regulatory compliance improved.
- [Pharma] Worked with C-suite stakeholders on Big Data and AI strategy, ranking use cases by business value, risk, and feasibility.
- [Pharma] Managed 42 client stakeholders and 14 Deloitte subject matter experts to generate use cases.
- [Financial services] Reviewed 50+ business processes and 500+ system controls, found the gaps, and recommended new controls.
Expand+
Jul 2017 – Aug 2019
Deloitte ConsultingNew York City
Business Technology Analyst
Data analytics and cloud migration programs for enterprise clients.
- [Financial services] Managed the migration of an automated analytics platform to AWS, used for account planning, pipeline management, and forecasting.
- Received Deloitte's Applause Award.
- Worked between business stakeholders and engineering, turning requirements into technical design documents.
Expand+
Oct 2014 – Aug 2015
KPMG IndiaGurgaon
Trainee
IT risk advisory: security architecture reviews, gap analysis, and ISO 27001 controls.
- Reviewed client security architecture and ran a gap analysis to find areas to improve.
- Wrote IT security policies and put controls in place to support ISO 27001.
- Coordinated a client's internal audit, which helped it get ISO 27001 re-certification.
Expand+
04 / Contact
Let's talk shop.
I'm open to product roles, and happy to compare notes on AI products or developer docs. Email is the quickest way to reach me.