AI Product Manager — practical AI systems, analytics, automation, and product execution.

CS background, William & Mary MBA in Business Analytics, and a bias for working software. I turn messy business workflows into AI products people can test, trust, and operate: define the use case, build the prototype, wire the data, measure quality, and decide what ships.

I work best at the point where product judgment meets hands-on AI execution. Give me an unclear workflow, a noisy dataset, or a half-formed automation idea, and I will turn it into a product-shaped system: scoped, tested, measurable, and explainable enough for another operator to run.

My lane is practical AI: n8n workflows, agents, analytics dashboards, RAG systems, evals, and product decisions that do not hide behind demo polish. The question I keep coming back to is simple: does this make the work faster, clearer, or easier to trust?

AI Automations n8n — 36 production-grade AI workflows. A practical n8n workflow collection for AI startup GTM and AI enablement for SMBs: lead generation, outbound, RAG, document intelligence, research workflows, support copilots, and internal ops. The point is not "I know n8n" — it is that I can map a messy business process, automate the repeatable parts, and leave a workflow another operator can understand. n8n · AI automation · RAG · document intelligence · GTM systems

AI Agents Autonomous — Agent patterns beyond chat demos. An agent-focused workspace for planning, tool use, workflow orchestration, and practical autonomous systems. I use this to explore where agents actually help: routing tasks, decomposing work, calling tools, keeping context, and handing a human a useful output instead of a vague answer. AI agents · autonomous workflows · tool use · orchestration · prompt systems

EvalForge — Stop shipping prompt changes on vibes. An open-source LLM regression-testing and release-decision tool for AI PMs and prompt owners. Compares two prompt/model variants against a golden dataset and returns a leaderboard, a category-level regression breakdown, and a decision memo with top risks and a recommended next action. Local, deterministic scoring — no paid APIs, no keys, no database. Python · Streamlit · LLM evaluation · release discipline   → Read the case study

StudyDesk — An assignment-first AI study workspace. A local-first desktop app that turns "I have 8 tabs open to write one paper" into a single workspace: three-panel assignment view, drag-to-cite from captured sources, a STEM toolbelt (calculator, formula book, similar-problem search), and one-click .docx export. Shipped as a tested MVP (400+ automated tests), built and reviewed across six incremental phases. Electron · React · TypeScript · local-first AI · AI agents · product shipping   → Read the case study

Instacart Basket Analysis — Turning order data into restocking decisions. SQL + Tableau + an interactive web dashboard analyzing Instacart order patterns to recommend smarter restocking, reorder timing, and inventory focus. The analyst muscle: from raw transactional data to a decision a category manager can act on. SQL · Tableau · product analytics · dashboarding · data storytelling   → Read the case study

LinkedIn-Codex — AI for the job search itself. A Codex-native LinkedIn job-review and resume-tailoring dashboard — an applied AI-automation build that scopes a real workflow and wires an LLM into it end to end. Python · AI automation · workflow tooling

Fraud Compliance Demo — FinTech risk command center. An end-to-end fraud and compliance workflow: transaction replay, risk scoring, deterministic rules, sanctions/PEP screening, live WebSocket dashboard, SAR draft artifacts, and executive fintech risk metrics. Built to show how raw transaction activity becomes analyst queues, audit evidence, and business-facing operating metrics. FinTech · fraud analytics · compliance workflow · FastAPI · Next.js

Healthcare Document Intelligence Hub — HealthTech document triage from synthetic notes. A PHI-safe document-intelligence pipeline for claims, prior authorization, care gaps, billing exceptions, and medication adherence. It extracts fields, classifies work, routes each item to an ops queue, scores urgency, and generates an executive dashboard from synthetic intake notes. HealthTech · document intelligence · workflow automation · Python · analytics

SaaS RevOps Command Center — Forecast risk, churn rescue, and next-best action. A synthetic revenue-operations pipeline that joins CRM, product usage, support, and GTM activity signals into account health, churn risk, expansion readiness, forecast category, and next-best action. It generates account-level JSON outputs plus a board-ready dashboard. SaaS · RevOps · GTM analytics · churn risk · Python · dashboarding

Retail Personalization Retention Engine — Customer lifecycle scoring and next-best offers. A synthetic ecommerce growth pipeline that joins customer profiles, orders, engagement signals, loyalty tiers, and category affinity to score churn risk, label lifecycle stage, recommend personalized offers, and estimate incremental revenue. Retail · ecommerce analytics · personalization · churn risk · Python · dashboarding

Supply Chain Risk Replenishment Engine — Stockout risk, supplier reliability, and reorder decisions. A synthetic operations analytics pipeline that joins SKU inventory, demand forecasts, open purchase orders, and supplier performance to score stockout risk, recommend reorder quantities, flag supplier risk, and generate an executive replenishment dashboard. Supply chain · logistics analytics · inventory risk · supplier reliability · Python · dashboarding

Cybersecurity GRC Alert Triage — Incident priority, control mapping, and evidence gaps. A synthetic security operations and GRC pipeline that joins alerts, asset criticality, MITRE tactics, controls, and evidence freshness to score incident risk, assign priority/SLA, map controls, and generate a CISO-ready dashboard. Cybersecurity · GRC · SOC2 controls · MITRE mapping · Python · dashboarding

HR Recruiting Ops Copilot — Candidate matching, funnel health, and outreach priority. A synthetic recruiting operations pipeline that joins jobs, candidate profiles, interview signals, and outreach history to score role fit, identify close-now candidates, flag offer/time-to-close risk, and generate a hiring dashboard. HR ops · recruiting analytics · candidate matching · funnel health · Python · dashboarding

Marketing Growth Attribution Engine — Attribution, budget moves, and lifecycle automation. A synthetic Growth Ops pipeline that joins campaign spend, web events, CRM pipeline, lifecycle automation, and content assets to score attribution, recommend budget shifts, and generate a marketing performance dashboard. Growth ops · marketing analytics · attribution · n8n automation · Python · dashboarding

EdTech Student Success Early Warning — Persistence risk, intervention routing, and advisor workload. A synthetic student-success analytics pipeline that joins LMS activity, assessments, support signals, and intervention capacity to score persistence risk, route next-best interventions, and generate an education operations dashboard. EdTech · student success · early warning · intervention analytics · Python · dashboarding

PropTech Portfolio Risk Optimizer — NOI, lease exposure, and asset-management decisions. A synthetic real estate analytics pipeline that joins property, financial, lease, maintenance, and market data to score portfolio risk, recommend stabilization or sale-review actions, and generate an owner-facing dashboard. PropTech · real estate analytics · NOI · lease risk · asset management · Python

InsurTech Claims Loss Ratio Copilot — Claims triage, reserve adequacy, and loss-ratio pressure. A synthetic insurance analytics pipeline that joins policies, claims, vendor performance, and portfolio thresholds to route claim priority, flag reserve/fraud/cycle risk, and generate a claims operations dashboard. InsurTech · claims analytics · loss ratio · reserve risk · fraud signals · Python

More: an interactive AI productivity course, and Lazy-docs, a collaborative editor built as an engineering take-home.

Product PRDs · RICE · PLG · GTM · requirement gathering · stakeholder communication · Agile/Scrum Analytics SQL · Tableau · Google Analytics · Amplitude · Power BI · dashboards · A/B testing Automation n8n · REST APIs · CRM automation · RAG workflows · GTM systems · support copilots Agents AI agents · tool use · task decomposition · orchestration · memory/context patterns AI LLM evaluation · prompt engineering · strict JSON schemas · Constraint Engine · Claude / Codex tooling Build Python · TypeScript · React · Streamlit · Electron · Git

Open to AI Product Manager, Product / Data Analyst, AI Analyst, AI Solutions & Automation, and Strategy & Operations roles. Available now. I am building and open-sourcing AI tools that show, rather than claim, how I work.

The fastest way to evaluate me: open AI Automations n8n, EvalForge, Fraud Compliance Demo, or InsurTech Claims Loss Ratio Copilot. That is the mix I bring: automate the workflow, evaluate the AI, and explain the analytics.

sunnysonimba26@gmail.com  ·  Portfolio  ·  LinkedIn