MVP Portfolio
Production-grade GenAI MVPs built with a reliability-first mindset: clear workflows, stable orchestration, and a stack that’s ready for real users.
Ship ad creatives faster—without trading off brand consistency
An AI pipeline that turns campaign inputs into ready-to-run video ad variants: scripts, voice, avatars, and reliable async rendering.
Queue design, orchestration, and reliability decisions—end to end.
Technical highlights
What makes this system stable under async workloads.
- Async rendering queue with retries + idempotent jobs
- Multi-variant script generation + structured prompt outputs
- Voice + avatar selection with clean fallbacks
- Stable state tracking across the generation pipeline
Production stack (grouped by architecture)
A stable default stack for shipping GenAI products: clear boundaries, predictable orchestration, and operational maturity.
Frontend
Product UI + rendering pipelines.
Backend
APIs, orchestration, background work.
AI / Media
LLMs, agents, voice, avatars, video.
View stack (8)
Infrastructure
Auth + data + deployment primitives.
More MVPs (case studies in progress)
These are real builds from recent work. The systems are already built and tested; documentation and deployment are pending, but they already follow the same production patterns: clear orchestration, stateful workflows, and reliability under async workloads.
Perplexity-style AI Research Tool
Full-stack AI research and search workspace for iterative investigations.
Agentic query routing with web search + SSE streaming, persisting sessions and research history in Supabase.
Capabilities
- Web search with citations + source tracking
- Multi-step reasoning with tool routing
- User sessions + saved research threads
- Structured research history + retrieval
AI Short Video Generator
Short-form video generation pipeline from prompt → script → voice → rendered MP4.
FastAPI orchestrates Gemini + media pipelines while Remotion workers render non-blocking jobs with outputs stored in Supabase.
Capabilities
- Deterministic MP4 rendering via Remotion workers
- Script generation + sequencing workflows
- TTS pipeline integration for voice tracks
- Async orchestration with retries + state
YouTube Analytics & Growth Workflow Tool
Authenticated creator analytics + creative workflow tool for repeatable growth ops.
Multi-step competitor analysis, ideation, and SEO workflows integrating Gemini and YouTube Data API in a Next.js + FastAPI stack.
Capabilities
- Competitor research + multi-step analysis
- SEO and performance insight workflows
- Creative ideation and planning assistance
- Thumbnail + creative asset workflow support
AI Logo Maker
AI branding tool for generating, iterating, and ranking logo variations.
LangGraph workflows drive generation loops, scoring, and iteration with user history persisted in Supabase.
Capabilities
- Prompt constraints + guardrails for style control
- Rank and compare variants via structured scoring
- Authenticated users + persistent logo history
- Iterative exploration of controlled variations
AI Interview & Candidate Scoring Platform
Voice interview platform for dynamic questioning and structured candidate scoring.
Real-time voice sessions via Vapi with stateful interview orchestration, transcripts, scoring, and recruiter dashboards.
Capabilities
- Low-latency voice interviews + live audio UX
- Dynamic question generation + response analysis
- Multi-dimensional scoring + rubric outputs
- Recruiter views + candidate session history
AI Medical Voice Agent
Voice-first medical intake assistant for triage, symptom capture, and scheduling.
Real-time clinical conversations via Vapi with structured extraction, follow-ups, and secure persistence in Supabase.
Capabilities
- Real-time voice intake with STT/TTS pipelines
- Structured medical data extraction
- Role-based access with authenticated workflows
- Audit-friendly state tracking + follow-ups
Engineering maturity (what you can rely on)
Reliability isn’t a feature—it’s the baseline. I build systems that behave predictably under async workloads and production constraints.
Reliability by default
Async-first workflows with retries, idempotency, and clear failure modes—so systems don’t silently drift in production.
Observability-minded
Structured logs, traceable job state, and measurable SLAs for the workflows that matter—especially long-running tasks.
Production constraints
Secure auth, sane data boundaries, and predictable orchestration so you can ship fast without rewrites later.
How I Build AI MVPs
A calm, production-first build process designed for founders who want speed without chaos.
- 1
Scope & System Design (48 hours)
Clarify the user journey, define critical workflows, and lock architecture decisions so execution stays fast and stable.
Deliverables
Architecture diagramAPI contractData modelRisk register - 2
Rapid MVP Build (2–4 weeks)
Ship the end-to-end product loop: auth, data, orchestration, UI polish, and production-grade failure handling.
Deliverables
Orchestration flowsJob state modelAdmin toolingCore UI screens - 3
Launch, Feedback, Iteration
Deploy, learn from real usage, and iterate with a tight loop—without rebuilding core systems each time.
Deliverables
Deploy pipelineMonitoring checklistPerformance passIteration roadmap