Shaping ideas into experiences… Driven by Insight. Designed for Impact.
6+ years scaling products at Microsoft (Excel Copilot) and Amazon (SDE). Currently architecting autonomous AI agents, LLM pipelines, and full-stack applications.
Sukul Bagai
Product leader & AI builder
❯whoami
Product Manager & AI Builder — ex-Amazon, ex-Microsoft
❯cat career.log
Amazon (SDE) → Microsoft (PM2, Excel Copilot) → sabbatical → AI builder, now
❯cat currently_building.log
· WhatsApp AI sales agent — live
· Algorithmic LLM trading agent — live
· AI-powered D2C brand — live
❯play video-resume.mp4
Video resume — coming soon
01 / Big Tech track record
Six-plus years of measurable outcomes across Microsoft and Amazon. Each company opens into a full PM case study — context, role, trade-offs, and impact.
Microsoft
Product Manager 2 · Excel mobile & iPad
98.9%+
cloud file-open reliability
1 mo → 1 day
feedback-triage time via LLM tool
+10M
MAU target funded for Mobile Copilot
#2 global
Microsoft Global Hackathon
Amazon
Software Development Engineer · Mobile Marketing
500K
DAU on ML notification pipelines
4×
order-rate lift from recommendations
+41%
sign-ins from notification pipelines
10M+
weekly users on App Banner platform
+15%
organic app installs
70%
query-latency reduction
02 / AI portfolio & labs
Three shipped AI projects, plus the concrete AI-PM artifacts — evals, cost models, PRDs — that back up the positioning instead of just claiming it.
Shipped
01 / whatsapp-sales-agent
Autonomous real-estate lead qualifier & scheduling engine.
Live demo coming soon
Embedded WhatsApp UI simulator replaying a live lead-qualification conversation.
02 / llm-trading-agent
Event-driven NIFTY options execution engine using market indicators.
Delivered ₹2L–3L profit during an active execution run.
View case studysoonLive demo coming soon
Architecture diagram + results visualization of the execution pipeline.
03 / litpartyshit
Direct-to-consumer brand powered by AI marketing-creative pipelines.
Live demo coming soon
Creative-pipeline showcase: from prompt to shipped ad asset.
In progress
Being fluent in AI systems means showing the artifacts, not just claiming the skill. Tracked openly as it gets built.
A live, working AI product
A small deployed tool anyone can open via URL — e.g. a PRD Review Buddy or a Voice-of-Customer clusterer, built as a simple RAG prototype. Ships with the system prompt, at least three iteration versions based on feedback, and real usage numbers documented alongside it.
The highest-signal artifact — recruiters at AI-first companies want direct evidence of shipping, not buzzwords.
An eval suite
A documented eval board for a real AI feature — roughly 30 happy-path, 10 edge, and 10 adversarial (prompt-injection) cases, each scored on independent dimensions: factuality, helpfulness, format adherence, refusal appropriateness. Published as a spreadsheet/JSONL with an experiment comparison across prompt variants.
Repeatedly called the single most under-prepared AI-PM skill — this proves it instead of claiming it.
An AI-native PRD
A one-feature PRD covering the six AI-specific sections: model-choice rationale (cost/latency/quality/vendor-risk trade-offs), the actual structured prompt, eval criteria, a cost model, failure modes with detection + response, and a human-in-the-loop strategy gating autonomy.
Most PRDs stop at the feature — this one shows the underlying system got designed.
A cost / pricing model
A spreadsheet modeling tokens per request (p50/p95, measured from real API calls), cost per request, monthly cost at 100/1k/10k users, caching impact, and margin analysis across free/pro/enterprise tiers.
Most AI PMs can't do this confidently — it's a concrete, rare differentiator.
A side-by-side model comparison
One real PM task — e.g. summarizing a customer-interview transcript — run across Claude, ChatGPT, and Gemini, compared on length, accuracy, structure, refusal behavior, and instruction-following, with a clear recommendation.
Shows the judgment to pick the right model for the job, not just familiarity with one.
A build-in-public log
Weekly progress write-ups (200–400 words: problem → approach → eval results → trade-offs → reflection) posted alongside each artifact above as it gets built.
Often generates inbound recruiter interest before the portfolio piece is even finished.
03 / Why I build
After 4.5 years PM'ing at Microsoft, I realized —
You cannot build exceptional AI products if you don't understand the underlying tech natively.
I took an intentional sabbatical to get my hands dirty — engineering agents, configuring vector databases, and managing token economics.
Today, I combine Big Tech product judgment with the speed and technical capability of a 0-to-1 AI builder.
Career journey
Latest first — building AI full-time, after a deliberate sabbatical from a Microsoft product career that started with an MBA and an engineering role.
Now
Independent AI Builder
Architecting autonomous agents, LLM pipelines, and full-stack AI products — the live projects in the AI Portfolio above.
TODO — confirm start date
Sabbatical
Stepped back from full-time roles to build hands-on AI systems.
TODO — confirm exact dates (~4.5 years)
Product Manager 2 · Microsoft, Excel mobile & iPad / Excel Copilot
Product roles at Microsoft, culminating in owning the Excel mobile & iPad Copilot experience.
TODO — confirm years
MBA · TODO — which program?
Went back to business school before moving into product management.
TODO — confirm exact dates (~2 years)
Software Development Engineer · Amazon, Mobile Marketing
Shipped the ML recommendation pipelines and Custom App Banner platform detailed in the track record above.
2017
Graduated · TODO — which college / degree?
Completed his undergraduate degree.
TODO — confirm years
Engineering internships · TODO — which companies?
Internships during college, before graduating.
04 / Skills
The overlap is the point: product strategy informed by hands-on time with the underlying systems.
01 · Product
02 · AI / LLM
03 · Engineering
04 · Tools
05 / Testimonials
Gathering testimonials — check back soon.
Worked with Sukul and want to share a few words? Reach out at hello@sukulbagai.com.
06 / Beyond work
Music, movement, and community — the parts of the story that don't fit on a résumé.
Music
DJ
One-line description — Sukul to add.
Movement art
Flow Art
One-line description — Sukul to add.
Visual tech art
Projection Mapping
One-line description — Sukul to add.
Community & education
Teaching
Community teaching work, including U&I and XLRI — description to add.
Animal welfare
PawSeva
One-line description — Sukul to add.
Side project
Live Mixxr
One-line description — Sukul to add.
Side project
Pips
One-line description — Sukul to add.
Also behind Stickman Ideates — a smaller, ongoing creative project.
07 / Contact
Open to Lead / Senior AI PM & Product Builder roles. If you're hiring — or building something hard with AI — the fastest way to reach me is email.