LinkedIn  ·  DealRith — live  ·  Design Catcher — live  ·  JobRith — live  ·  Montréal, Canada

I’m Yash, an AI engineer in Montréal with roots in Gujarat. I like the part of engineering where the problem is still messy: sitting close to users, finding the real constraint, building the smallest useful proof, and carrying it all the way into production.

My work spans agentic AI, LLM orchestration, product engineering, cloud systems, and observability. I’m most useful when a team needs one person to connect customer context, system architecture, and hands-on delivery.

How I work

• Discover the actual problem. Talk to users, map the workflow, and separate the painful constraint from the requested feature.
• Build to learn. Prototype the riskiest path first, connect real data, and evaluate behavior before scaling the surface area.
• Deploy responsibly. Add authentication, observability, cost controls, failure paths, and feedback loops so the system survives contact with real users.

Production systems and open tools

DealRith / Honeycomb — acquisition intelligence from discovery to close

DealRith brings business-for-sale discovery, broker and proprietary data, wishlists, documents, messages, and pipeline operations into one acquisition workspace. The wider Honeycomb system spans a React/TypeScript frontend, Java/Spring services, Python data collection, Firebase authentication, MongoDB, DynamoDB, API gateways, and an eight-stage deal process from teaser and NDA through LOI and under contract.

Availability: Live product → · Private multi-repository implementation

React · TypeScript · Java · Spring Boot · Python · Firebase · MongoDB · DynamoDB

Design Catcher — turn a live website into design context for AI agents

Design Catcher extracts DOM structure, computed styles, design tokens, layout, and motion from a live website, then synthesizes DESIGN.md, STRUCTURE.md, and UI-KIT.md files that coding agents can actually use. Its architecture combines a Chrome MV3 extension, deterministic browser extraction, Gemini on Vertex AI, Firebase caching and Cloud Functions, GitHub synchronization, and an Astro product site.

Availability: Live tool → · Public capture gallery → · Private product implementation

WXT · React · TypeScript · Chrome MV3 · Vertex AI · Gemini · Firebase · Astro

Synaptic — a voice-first second brain

Synaptic turns spoken thoughts into structured notes, todos, goals, meetings, and projects, then connects them through a user-owned knowledge graph so AI agents retain useful context. I designed it as a hybrid mobile-cloud system with a Flutter client, Java/Spring Boot services, Firestore, Vertex AI, streaming responses, offline fallbacks, and MCP integrations.

Availability: Live product → · Architecture paper → · Private product implementation

Flutter · Java 21 · Spring Boot · Vertex AI · Firestore · Cloud Run · MCP

JobRith — an AI workspace for the job search

JobRith brings job discovery, fit evaluation, tailored application material, outreach, and pipeline tracking into one workspace. The system combines a React/TypeScript product surface with Python services, browser automation, LLM workflows, Firebase, Cloud Run, and a Cloudflare edge layer.

Availability: Live product → · Private product implementation

React · TypeScript · Python · Flask · LLM workflows · Cloud Run · Cloudflare

ChainSight — supply-chain risk intelligence

ChainSight ingests external news and GDELT signals, applies NLP-based sentiment and risk analysis, and surfaces developing supplier threats through dashboards and alerts. It is a multi-service system spanning Java, Python, TypeScript, Firebase, data pipelines, and model-backed analysis.

Availability: Private research system

Java · Python · TypeScript · NLP · GDELT · Firebase · Microservices

Visual Vibes — architecture diagrams grounded in the code

Visual Vibes is an agent skill that inspects a repository once, builds an evidence-backed system model, and generates consistent C4, UML, ERD, DFD, sequence, deployment, and security views. It solves a familiar documentation failure: diagrams that look polished but drift away from the implementation.

Availability: Public source →

TypeScript · Agent skills · Mermaid · draw.io · Architecture analysis

Engineering range

• AI systems: Vertex AI · Gemini · OpenAI · Claude · MCP · RAG · agent workflows · evaluation
• Product engineering: Python · TypeScript · Java · Dart · React · Flutter · Spring Boot · Flask
• Cloud and operations: Google Cloud · Firebase · Cloudflare · Docker · Redis · OpenTelemetry

Open work and field notes

• AI Engineering Roadmap — a practical 24-week, project-led path through ML, deep learning, generative AI, agents, evaluation, and responsible AI.
• Demo Video Master — an agent workflow that turns a live product feature into a cinematic demo using browser automation, Remotion, and generated narration.
• Git Repo to Video — a developer tool that translates a repository into a concise, visual product story.
• Claude Landing Page Studio — an agent skill for researching, designing, building, and improving conversion-focused landing pages with strong SEO foundations.

Let’s build something useful

If you’re working on an AI product that has to cross the gap between a promising demo and a dependable user workflow, I’d like to hear about it.

Start a conversation on LinkedIn →
Explore my repositories →
Watch Code to Develop →

_{Built in Montréal. Informed by Gujarat. Always close to the problem.}