Hey, I'm Haroon Riasat.
4.5 years taking web, mobile and AI products from idea to production: RAG assistants, transcription pipelines and document extraction, on React, Python, Node.js and AWS.
Product teams, a consultancy I co-founded, and a team I led. Most client work is under NDA, so the details live here, not on GitHub.
Client products I led, plus side projects you can open right now.
Transcribes long, noisy Maui County court-hearing recordings with Deepgram and uses Claude to turn them into structured, lawyer-ready notes. I led the architecture and delivery.
Family health platform: lab-report PDFs and photos become structured results with trends, and an MCP server lets Claude use it with role-scoped tools and human approval.
AI-powered CRM forms and reporting that automates field inspections, checklists and report generation. I led development and owned the core build and AWS deployment.
A RAG chatbot grounded only in my profile: hybrid vector and keyword retrieval, cited sources, and guardrails that say "I don't know" instead of guessing.
A custom sales-management app extending Salesforce, with a React/TypeScript frontend, Express/PostgreSQL backend and OAuth sign-in.
A live animal rock concert in the browser. Upload your own track and the band learns it, or host a party so friends join from their phones in sync.
Every tool here comes with where I used it. Levels follow my resume: what I use daily, what I've shipped, and what I've worked with.
OpenAI, Claude and Gemini. OpenAI in Amplon's RAG assistant and Subqc AI, Claude in Council Tap and Kinwell.
Embeddings with pgvector, Pinecone and Upstash Vector. Amplon's in-product assistant and my AI resume.
Fine-tuned LLMs on client data at QUEST Lab so responses fit each client's domain.
A Node.js MCP server for Kinwell, with tools scoped to each user's role and human approval before messages go out.
Speech-to-text for long, noisy court-hearing recordings in Council Tap.
Orchestration and retrieval frameworks for LLM features.
Claude, GitHub Copilot and ChatGPT in daily coding, review and debugging.
Amplon's Hoshin Kanri platform and most Exitech client products.
Frontend and Node.js work, SCRUB's Salesforce extension and my AI resume.
Headless CMS integrated with Amplon's React frontend for release notes and in-app content.
Native iOS apps alongside web products at QUEST Lab.
Flask powers Amplon's RAG assistant; Python also runs Council Tap's transcription pipeline.
Exitech client backends, SCRUB's Express API and Kinwell's MCP server.
Backend features and APIs at Amplon, and an internal audit system during my QUEST Lab internship.
REST APIs, OAuth and a Salesforce extension (SCRUB); Contentful and MCP integrations.
Exitech client products and SCRUB's backend.
MERN products: Subqc AI and Council Tap.
EC2, S3, Amplify, Bedrock and SageMaker. Deployments at Exitech, QUEST Lab and Amplon.
Containerised AWS deployments for Exitech clients.
Pipelines from commit to production for Exitech products.
Paste a job ad. My AI compares it with my resume, requirement by requirement, and is honest about the gaps too.
Your match report appears here: strengths, partial matches and gaps, each with evidence from my resume.
Scalable Software Systems. Thesis: TaskLint, an AI Chrome extension that audits project-board tasks for clarity, ownership, deadlines and delivery risk.
Dean's List of Honor three times. Chair of the IEEE Computer Society chapter, with 80+ tech events organised.
I'm based in Turku and work in English. Email is the fastest way to reach me, or ask my AI resume anything first.