AJ OTRANTO
AJ OTRANTO
Not a Developer. Not an Agency. Something More Useful.
I help companies figure out where AI actually belongs in their product—and then I build it with them. No decks. No "discovery phases." Just working software, shipped fast.
January 2026
By day, I work in advertising—10 years of media strategy, product launches, and understanding what makes people actually use things. On the side, I build AI products. Started in 2020, before the hype cycle, before the gold rush.
Here's what I've learned: AI doesn't have to be a massive initiative. The best use of it right now is solving specific business problems that traditionally required buying enterprise software—only to use 1% of what you paid for. I build just that 1%, for a fraction of the cost.
I work directly with founders and product teams to look internally at what's actually slowing you down, then build bespoke solutions that fit your exact workflow. Production-ready in days, not quarters. No handoffs. No sprint cycles. Whiteboard to deployed.
What I'm not: A staff augmentation play. A consultant who disappears after the PDF. An agency that needs managing. I work embedded, accountable, and fast.
"Most AI projects fail because they start with 'what can AI do?' instead of 'what's actually broken?'"— AJ Otranto
THE DESK LEDGER
The Process
CASE FILES
What I built, how I built it, and what I learned
See Your Wedding Before You Book It
Production-grade SaaS for venue visualization. Multi-tenant architecture, Stripe billing, real-time collaboration, AI generation engine with custom prompt engineering. My first major build—learned what persistence and determination actually mean when shipping.
Next.js, Supabase, Stripe, Resend, Apollo, Gemini, Claude
View project →Enterprise Tools Under NDA
Multi-module platform for media planning teams: AI brainstorming with fit scoring, influencer tracking with budget guardrails, node-based media planning, and forecast modeling. Key learning: embed with subject matter experts—make them feel ownership of the build.
Next.js 14, Supabase RLS, Gemini, OpenAI, Recharts, dnd-kit
AI Dance Choreography Generator
Weekend project for a colleague's daughter—generate dance moves synced to her favorite TV show songs. Spotify audio analysis (tempo, energy, beats) + LRCLIB synced lyrics + Gemini image generation. Includes a DDR-style rhythm game. Learned: the magic happens when you orchestrate APIs in unexpected ways.
Next.js, Spotify Web Playback SDK, LRCLIB, Gemini Vision
The "Everything App" That Taught Simplicity
Attempted to build the perfect ADHD note-taking platform—Tiptap editor, workspaces, AI chat, voice input, auto-tagging, task extraction, note merging. Scope crept for months. The lesson: sometimes less architecture is more.
Next.js, Tiptap, Supabase, Claude AI, Voice API
This Site Is the Portfolio
A newspaper-themed portfolio that's also a design system testbed. Each version experiments with interaction patterns, animation timing, and information hierarchy. Proof that good taste and fast execution aren't mutually exclusive.
Next.js 14, Tailwind, Framer Motion
View project →CLASSIFIEDS
PROFESSIONAL SERVICESRates available upon inquiry. Click listing for details.
AI Product Consulting
Figure out where AI belongs—and where it doesn't
Most AI features fail because they solve the wrong problem. I help teams identify the highest-leverage opportunities for AI in their product, then map out exactly how to build them. No fluff, no "AI strategy deck"—just clarity on what to build and why.
INCLUDES:
- • Product audit and opportunity mapping
- • Technical feasibility assessment
- • Build vs. buy recommendations
- • Prioritized feature roadmap
- • Ongoing async support
click for details
Build Partnership
Embedded building, from idea to production
I work directly with your team to design, build, and ship AI features. Not a handoff—a partnership. You get a product thinker and builder in one, moving at startup speed inside your organization.
INCLUDES:
- • Feature design and architecture
- • Production-ready implementation
- • Direct collaboration with your team
- • Deployment and iteration support
- • Knowledge transfer and documentation
click for details
Advisor Mode
Ongoing access for teams building with AI
For teams that have the builders but need a technical sounding board. Weekly calls, async access, and someone who's seen the patterns before. Think of it as having a fractional AI product lead on speed dial.
INCLUDES:
- • Weekly strategy calls
- • Async Slack/email access
- • Architecture and code review
- • Vendor and tool evaluation
- • Team coaching and unblocking
click for details
To inquire about rates or availability, contact: ajotranto@gmail.com
WHERE AI BELONGS
Three modes of AI that actually work in products
Consultative
AI that advises but doesn't decide. Think recommendations, summaries, suggestions. The human stays in control.
E.g., "Based on your history, you might like..."
Decisive
AI that makes choices within defined bounds. Automating decisions humans don't want to make or can't make fast enough.
E.g., Auto-categorizing support tickets, fraud detection
Invisible
AI that works behind the scenes, improving experiences without users knowing it's there. The best kind.
E.g., Search ranking, content moderation, personalization
The best AI features fit into one of these three categories. If your feature doesn't fit cleanly, it's probably trying to do too much. Clarity of purpose is the first sign of good AI product design.
ASK THE DESK
Get AI product advice, brainstorm ideas, or ask about my work
This isn't a generic chatbot. It's trained on my perspective, my opinions, and my experience building AI products. Think of it as a quick way to get the kind of advice I'd give in a real conversation.
Three modes to choose from:
- Consult: "Should I use RAG or fine-tuning?" Get opinionated AI product advice.
- Ideate: "AI ideas for a wedding app..." Brainstorm features and solutions.
- Ask AJ: "What's your experience with X?" Questions about my work.
AI Assistant
Get AI product advice — should you build it, buy it, or skip it?
Ask me anything in Consult mode
FIELD NOTES
Observations from the front lines of AI product building
The 80/20 of Prompting
Jan 2026Most prompting advice is noise. What works: be specific about format, give examples, and tell the model what NOT to do. That's it.
When to Use Fine-Tuning (Rarely)
Jan 2026Fine-tuning is usually a sign you haven't tried hard enough with prompting. It's expensive, slow, and locks you into one model. Use it only when you have proprietary data that fundamentally changes task performance.
The "AI Feature" Trap
Dec 2025Adding AI to a feature doesn't make it better. Most products would be improved by removing AI, not adding it. The question isn't "can we use AI here?" but "should we?"
Speed Over Polish
Dec 2025A working prototype in a week beats a perfect spec in a month. Real users teach you things documents never will. Ship ugly, learn fast, then polish.
The Latency Tax
Nov 2025Users will tolerate slow AI responses exactly once. After that, they'll use something faster. Optimize for perceived speed: streaming, skeleton loaders, progressive disclosure.
Build for the Skeptic
Nov 2025Design your AI features for the user who doesn't trust AI. Give them transparency, control, and escape hatches. If you convince the skeptic, you'll delight everyone else.
What are you trying to solve?
Tell me about the challenge. What's the problem? What have you tried?
Step 1 of 3
THE WEEKLY STRIP
"Just Add AI" - A comic about building in the AI era
PM: "Can we add AI to the search?"
Dev: "It already has autocomplete..."
PM: "Right, but can we make it...smarter?"
Dev: *opens ChatGPT to ask what that means*
"Just Add AI" by AJ Otranto • Episode 1: The Feature Request
New strip every Friday
TECH HOROSCOPES
For entertainment only*Your weekly fortune in the age of AI
♈ Aries
Your next pull request will be approved on the first try. The stars are aligned, but your linter is not.
♉ Taurus
Resist the urge to rewrite everything in Rust this week. Sometimes JavaScript is just fine.
♊ Gemini
You will discover a bug that has been in production for 3 years. No one will believe you.
♋ Cancer
Mercury is in retrograde, which is why your API keeps returning 500 errors.
♌ Leo
Your demo will work perfectly in development and crash spectacularly in front of stakeholders.
♍ Virgo
This is the week you finally read that documentation. Just kidding. Stack Overflow it is.
♎ Libra
A framework you love will be deprecated. Mourn briefly, then migrate immediately.
♏ Scorpio
Your AI model will hallucinate something profound. Write it down before you forget.
*Accuracy not guaranteed. Past performance does not predict future deployments. Not financial or technical advice.
HUMAN INTEREST
The person behind the products
When I'm not building AI products, I'm usually doing one of three things: trying to stay consistent in the gym, getting humbled by Spanish, or disappearing into some rabbit hole about design systems and why "simple" products are almost never simple.
A lot of my free time looks like small experiments. I'll build a tiny tool to solve an annoying problem, obsess over the workflow for a week, then either ship it, scrap it, or fold it into the next idea. I'm drawn to things that remove friction, especially the kind you don't notice until it's gone.
I've also been a chess person for as long as I can remember. My poppy taught me how to play when I was young, and it's something I've stuck with ever since. Not in a "serious tournament arc" way, more in a "give me puzzles, openings, and a quiet hour to think" way. I like the discipline of it: pattern recognition, trade-offs, and learning by being wrong in public (or at least on the board).
I live in New York (Williamsburg), where the rent is objectively insane but the city still feels like the best place to build, learn, and keep your standards high. I'm engaged, and most weeks are some mix of work, building, seeing friends, family time, and trying to be a normal person who doesn't turn everything into a project.
Fun fact: my greatest accomplishment is getting on the wall at my local Applebee's (Floyd Football, 2013). Peak performance.
THE DAILY CROSSWORD
Interactive puzzleTest your knowledge of my world—AI, advertising, chess, and Spanish
Across
- 1. LLM input that shapes the output (6)
- 7. Madison Ave. abbreviation (2)
- 8. Castle in chess, but castles in romance (4)
- 10. "Hello" south of the border (4)
- 11. Audience you aim for (6)
- 12. Transformer component that outputs text (7)
Down
- 1. Chess piece that opens the game; key AI term for context (4)
- 2. AI that generates images, not the beverage brand (5)
- 3. Temperature setting for predictable AI output (6)
- 4. Artificial ___ (abbr.) + Spanish goodbye = ad strategy (5)
- 5. David Ogilvy's old agency initials now in WPP (5)
Theme: The AI x Advertising Intersection
A cryptic mix of AI terminology, advertising jargon, chess pieces, and Spanish—where my worlds collide. Clues may be wordplay, double meanings, or industry-specific. Good luck.