Case Studies
Real projects, real architecture decisions, real outcomes. How we built AI products for startups — and what the results looked like.
22 articles
How We Built the AI Data Analyst: Plain English to PDF
Build story: plain English → SQL → chart → board-ready PDF. Three weeks, GPT-4o, Plotly, and the chart-type bug we caught at the client demo.
Abraham Jeron
How We Built an AI Education Content Creator
4-week build story: AI pipeline that cut EdTech course production from 3-4 weeks to 1 day. What broke, how we fixed Bloom's calibration, real numbers.
Abraham Jeron
How We Built SARA: Speech-to-Action, Not a Booking Bot
What we learned building SARA, a voice AI agent for internal ops. Why speech-to-action differs from booking bots, and when it's worth building.
Abraham Jeron
How We Built the Content Engine That Powers Fertilia Health
From 0 to 5,000 weekly Google impressions in 5 weeks, then 50,000+ weekly impressions in another 5 weeks (9× growth). The architecture, the wrong turns, and the physician-review queue that made it work.
Abraham Jeron
Meeting Intelligence Tool: 4-Hour Recording to 1-Page Brief
How we built a meeting intelligence tool that converts 4-hour recordings to a 1-page brief. Semantic chunking, multimodal frames, and what broke.
Abraham Jeron
How We Built Our Call Compliance AI: 5 Decisions
5 decisions that shaped our sales call compliance AI, from rubric design to dashboard UX. What the client said in our retrospective.
Abraham Jeron
How We Built an AI Workflow Automation System with n8n
How we replaced 4 hours of daily manual processing with n8n and GPT-4o. Architecture, failures, and real numbers after 6 weeks in production.
Abraham Jeron
How We Built an Advanced RAG System for Documents
Build story: 900 contracts turned into a queryable AI system. Naive RAG got 0.61 precision. Hybrid search + reranking hit 0.87. What we rebuilt and why.
Abraham Jeron
How We Built a Talent Matching Platform with AI
Build story: AI matching for an entertainment talent platform. Semantic search with pgvector, profile normalization, and what we got wrong first.
Abraham Jeron
Speech-to-Action: How We Built the SARA Speech Agent
Real build story: SARA converts speech to app actions in under 2 seconds. WebSocket streaming, Deepgram Nova-2, intent classification, and what broke first.
Abraham Jeron
Form Automation with AI: Cutting 40 Hours of Data Entry
How we automated 40 hours of weekly form data entry using AWS Textract and GPT-4o. Architecture, validation logic, and the exception threshold question.
Abraham Jeron
How We Built a Coding Assessment Tool with AI Evaluation
How we built a coding platform with a custom compiler engine and AI evaluation. What broke with OneCompiler, how we scored code approaches, real build story.
Abraham Jeron
How We Built a Unified Workspace with AI-Powered Search
How we built AI-powered search across Slack, Notion, Google Drive, and email. RAG architecture, connector headaches, and real latency numbers.
Abraham Jeron
Video Auditing with AI: Building an Automated QA Pipeline
Real build story: AI video auditing system with frame sampling, vision models, and checklist rubrics. What broke, what worked, real accuracy numbers.
Abraham Jeron
AI for EdTech: Course Dev from 4 Weeks to 1 Day
How an EdTech provider cut course development time by 95% using AI. The pipeline, the architecture, what broke first, and the real numbers.
Venkataraghulan V
How We Built a Trading Analytics Platform with Real-Time AI
How we split hot path from AI path in a real trading platform, added NL queries on live data, and handled the latency decisions that changed the architecture.
Abraham Jeron
How We Built AI-Powered Evaluation for an EdTech Platform
Real build story: AI rubric extraction and semantic scoring for an EdTech assessment platform. What broke, what worked, real accuracy numbers.
Abraham Jeron
Building a Meeting Intelligence Tool: Full Architecture
How we built a meeting intelligence tool with transcription, speaker diarization, action item extraction, and searchable history. Architecture decisions and tradeoffs.
Abraham Jeron
How We Publish 14 Blog Posts a Week Using AI Agents
Our real AI content pipeline: automated drafting, human review, SEO feedback loops, and LinkedIn distribution. The full technical stack behind 14 posts/week.
Abraham Jeron
What We Learned Building an AI Call Analyzer
What we learned building an AI call analyzer processing hundreds of sales calls daily. Architecture, cost per call, and when to build vs buy Gong.
Abraham Jeron
How We Built a Data Analyst You Can Talk To
Real build story: text-to-SQL AI for non-technical finance users. Schema-aware prompting, query validation, reasoning panel. Real accuracy numbers.
Abraham Jeron
How We Built a Sales Call Compliance AI in 2 Weeks
How we scored 94% agreement with human reviewers using Claude on diarized sales call transcripts. Deepgram + speaker diarization + rubric scoring. Architecture and code.
Abraham Jeron