Things I've Built

Products built with AI as a pair-builder.

Side projects designed, coded, and shipped in close collaboration with AI: human judgment on what to build, AI leverage on how fast it ships.

✦ Each of these was built as a human + AI collaboration, with Claude as a pair-builder across design, code, and content. Giving credit where it's due.

SwingScope

Guided, educational market analysis for swing traders

A demo-ready MVP that helps intermediate swing traders learn why a setup is or isn't forming. Pairs a React + TypeScript frontend with a FastAPI backend to deliver guided technical analysis, without ever presenting financial advice.

Search US tickers and build a single master watchlist with delayed price data and candlestick charts.
Deterministic technical-analysis strategy engines explain the conditions behind each setup.
AI panel with Technical Analyst, Research Assistant, and Learn modes, plus annotated teaching moments directly on-chart.

ReactTypeScriptFastAPIPythonAI

Educational analysis platform, not a brokerage or financial advisor.

CrimeReel Studio

True crime cases → serialized short-form video, automatically

An AI-powered content pipeline that transforms real true crime cases into serialized short-form video scripts optimized for Instagram Reels, YouTube Shorts, and Facebook. Built for the @crimewithmasala brand, it automates case research, script generation, pacing analysis, and voiceover, cutting production time from hours to minutes.

Input a case via URL, pasted text, or auto-suggestion, and the app scrapes, summarizes, and generates a serialized 3-episode script tuned for 30-60 second Reels.
Every episode gets pacing feedback, word-count analysis, cliffhanger scoring, and cut suggestions.
Generates AI voiceover, sources dark cinematic B-roll, and assembles the final 9:16 vertical video, ready to download and post.

ReactViteFastAPIClaude APIElevenLabsPexels APIMoviePyVercelCloud Run

Coffee Quest

Learn espresso by feel, without burning beans

A single-page React app for people getting into specialty coffee: 38 honest equipment reviews, an interactive brewing simulator, and a persistent brew tracker with trend visualization. Zero npm dependencies: React 18, Tailwind, and Babel run straight from CDN.

The Brew Slider: four live controls (grind, temperature, ratio, time) drive real-time flavor verdicts that teach the cause-and-effect loop.
Filterable catalog of 38 machines, grinders, and accessories from an $8 spray bottle to an $8,495 La Marzocco.
Brew log saved to localStorage with a hand-rolled inline-SVG trend chart: no chart library, ~50 lines.

React 18TailwindInline SVGlocalStorageZero-build

Consulting Case Studies

Selected client work.

Voice-Enabled Market Insights Assistant

Designed and delivered an agentic AI assistant integrating LLMs, RAG, AlphaSense, and Factiva, from concept to adoption.

Adopted by a large internal market insights team and actively being integrated into a second platform.
Combined voice interaction, retrieval workflows, and external market intelligence sources into one research flow.
Built the processes and onboarding needed to scale beyond the initial pilot.

LLMsRAGVoice AIAPI Integrations

Consumer Communications Root-Cause Analysis

SQL-driven analysis across millions of records to isolate duplicate consumer notice patterns and coordinate the full fix cycle.

Mapped the operational and system conditions creating redundant outreach.
Coordinated design through UAT and go-live for the fix.
Reduced redundant outreach by 20% with zero recurrence in six months post-deployment.

SQLAnalyticsOperations

Pricing & Usage Analytics at Badger Maps

Product analytics and pricing modeling that guided rollout decisions and revenue improvements.

Analyzed 200K+ user records in PostgreSQL to quantify route optimization value (~$71 saved per user).
Python engagement analysis drove a 30% increase in route optimization usage.
R pricing simulation backed a 5% adjustment that improved retention.

PythonRPostgreSQLExperimentation