Four case studies with the problem, my role, the stack and the outcome. Then the full index, filterable by the kind of work. Every entry links to the live product.
All
10
AI and ML
07
Full stack
06
Backend
06
Frontend
04
Case studies · 0401 / 04
01 / Nysonian
Nysonik
Problem
Customer messages arrive by email, chat, Instagram, voice and Shopify. Without one place to answer them, the context of a conversation and its response time split across five tools.
Role
Full Stack AI Engineer at Nysonian. AI features over the customer record store, live updates over WebSockets, JWT authentication and the Shopify connection.
03 / Silverthread Labs
Ezly
Problem
People who sell for a living write the same kinds of replies all day, across several apps, and every one has to sound like them. Generic model output reads as generic.
Role
Full Stack Developer at Silverthread Labs. Model inference pipelines, persona based generation, saved replies, proposal generation and the full Next.js and TypeScript interface.
04 / Devyst
DevflowAI
Problem
Standups eat the morning, and new engineers spend their first weeks asking where things live. The repository and the team's written updates already hold most of those answers.
Role
Designed and wrote it end to end. Open source and self hosted, with users bringing their own model keys.
05 / Silverthread Labs
GlanceAI
Problem
Reading a long page for the three facts you need is slow. The page already holds the answer, it needs a reader that can pull it out and take follow up questions.
Role
Developer at Silverthread Labs on the extension and the web app: page summaries, key points, chat over page content and related links.
Fig 01 · Nysoniknobl.nysonik.com
Fig 03 · Ezlygetezly.com
Fig 04 · DevflowAIdevflowai-by-devyst.vercel.app
Fig 05 · GlanceAIglanceai.site
Stack
Next.js
TypeScript
Node.js
PostgreSQL
WebSockets
OpenAI API
Shopify
Outcome2.8M+conversations handledAlso published: 38% resolved by AI, 1m 42s median first response.
Customer messages arrive by email, chat, Instagram, voice and Shopify. Without one place to answer them, the context of a conversation and its response time split across five tools.
Role
Full Stack AI Engineer at Nysonian. AI features over the customer record store, live updates over WebSockets, JWT authentication and the Shopify connection.
Stack
Next.js
TypeScript
Node.js
PostgreSQL
WebSockets
OpenAI API
Shopify
Outcome2.8M+conversations handledAlso published: 38% resolved by AI, 1m 42s median first response.
People who sell for a living write the same kinds of replies all day, across several apps, and every one has to sound like them. Generic model output reads as generic.
Role
Full Stack Developer at Silverthread Labs. Model inference pipelines, persona based generation, saved replies, proposal generation and the full Next.js and TypeScript interface.
Stack
Next.js
TypeScript
Python
OpenAI API
Chrome Extension
Outcome400people drafting with it28 of them paying customers, per getezly.com.
Standups eat the morning, and new engineers spend their first weeks asking where things live. The repository and the team's written updates already hold most of those answers.
Role
Designed and wrote it end to end. Open source and self hosted, with users bringing their own model keys.
Stack
Next.js
TypeScript
Claude API
OpenAI embeddings
Semantic search
OutcomeOpen sourceself hosted, bring your own keys
Reading a long page for the three facts you need is slow. The page already holds the answer, it needs a reader that can pull it out and take follow up questions.
Role
Developer at Silverthread Labs on the extension and the web app: page summaries, key points, chat over page content and related links.
01One inbox for email, chat, Instagram, voice and Shopify threads.
02Role based access, audit trails and redaction for security conscious brands.
03AI assistants and realtime insights over the conversation history.
Problem
Customer messages arrive by email, chat, Instagram, voice and Shopify. Without one place to answer them, the context of a conversation and its response time split across five tools.
Role
Full Stack AI Engineer at Nysonian. AI features over the customer record store, live updates over WebSockets, JWT authentication and the Shopify connection.
03Meeting booking with calendar sync and handoff to account executives.
04Pipeline analytics, with imports from a CRM or CSV.
Problem
Sales development reps work a queue of prospects by phone and email, then hand qualified meetings to account executives. Those steps usually live in three separate tools.
Role
Full Stack AI Engineer at Nysonian, on the queue, sequence and analytics features.
01Personas set the voice. The user still sends every message.
02Reads the deal stage before it drafts a reply.
03Team seats with shared personas and admin controls.
Problem
People who sell for a living write the same kinds of replies all day, across several apps, and every one has to sound like them. Generic model output reads as generic.
Role
Full Stack Developer at Silverthread Labs. Model inference pipelines, persona based generation, saved replies, proposal generation and the full Next.js and TypeScript interface.
01Written updates roll into daily summaries and a weekly sprint digest by email.
02Connect a repository and the onboarding agent writes a guide to it.
03Grounded answers cite the exact files and line ranges they used.
Problem
Standups eat the morning, and new engineers spend their first weeks asking where things live. The repository and the team's written updates already hold most of those answers.
Role
Designed and wrote it end to end. Open source and self hosted, with users bringing their own model keys.
Stack
Next.js, TypeScript, Claude API, OpenAI embeddings, Semantic search
01Key points and a summary of the page that is open.
02Chat against the page content.
03Related results for further reading.
04Runs in Chrome, Brave and Opera.
Problem
Reading a long page for the three facts you need is slow. The page already holds the answer, it needs a reader that can pull it out and take follow up questions.
Role
Developer at Silverthread Labs on the extension and the web app: page summaries, key points, chat over page content and related links.
Stack
Chrome MV3, TypeScript, Next.js, LLM APIs, Service Workers
01Write, chat, translate and ask from the same panel.
02Text out of images with OCR.
03Grammar checks and reference generation.
Problem
Students and professionals switch between separate tools to write, translate, check grammar and cite sources, and every switch costs the thread of what they were doing.
Role
Extension and web app: the tool panel, OCR flow and the model calls behind each tool.