Technical Arsenal

Qaswar Sarfraz — Flutter, Dart & AI Engineering Skills | Twenty-nine tools.
Every one with receipts.

No self-assigned percentages here. Each technology below carries the tier I actually operate at and the product, role or credential that proves it — so you can verify the claim instead of taking my word for it.

0 Technologies
0 Production grade
0 Domains
0 In the arsenal
Evidence, not estimates

The arsenal, by domain

Filter by discipline. Every row names where the tool was used.

Production Shipped and running in a live product
Advanced Used substantially in real work or coursework
Working Building competence on a project in progress

Mobile Engineering

Cross-platform iOS & Android
07
  • Flutter Widget composition, custom painters, 60fps render budgets Production LectureMind · Tripline · CSOFT
  • Dart Null safety, isolates, async streams, extension methods Production LectureMind · Tripline · CSOFT
  • Riverpod Provider graphs, async notifiers, dependency scoping Production LectureMind · Tripline · CSOFT
  • GoRouter Declarative routes, auth guards, deep-link handling Production LectureMind · CSOFT
  • flutter_map & Geolocation Vector tiles, marker clustering, live position streams Production Tripline
  • Provider Legacy state layers and migration paths to Riverpod Advanced University projects
  • Android Studio & Gradle Build flavours, signing configs, release pipelines Advanced CSOFT Systems

AI & LLM Systems

Low-latency inference in mobile apps
06
  • Groq · Llama 3.3 70B & GPT-OSS-120B Sub-second inference, neural STT normalizer & JSON-constrained output Production LectureMind · Tripline
  • AssemblyAI · Universal-3.5 Pro Real-time WebSocket audio streaming, live STT, academic word boost Production LectureMind
  • Streaming REST & WebSocket inference Audio streaming, token duplexing, backpressure, cancel-safe requests Production LectureMind · Tripline
  • Prompt design & guardrails System contracts, schema validation, refusal handling Production LectureMind · Tripline
  • Cost & latency budgeting Token accounting, response caching, model selection Advanced LectureMind · Tripline
  • Document Knowledge Synthesis Slide text parsing, vector PDF export, Canvas mind map generation Production LectureMind

Cloud, Data & APIs

Postgres, row-level security, offline sync
06
  • Supabase Auth providers, storage buckets, realtime channels Production Tripline · CSOFT
  • PostgreSQL & RLS Schema design, row-level security policies, query tuning Production Tripline
  • Hive encrypted store Offline-first cache, encrypted boxes, schema migrations Production CSOFT Systems
  • REST APIs with Dio Interceptors, retry policy, typed error envelopes Production CSOFT Systems
  • Firebase & Secure Storage Mobile Ads SDK, token vaulting, crash reporting Advanced CSOFT Systems
  • FastAPI services Async endpoints, Pydantic contracts, dependency injection Working Document AI · in progress

Engineering Foundations

The part that outlives any framework
05
  • Object-oriented design SOLID, composition over inheritance, design patterns Production Parking System
  • Java Core Java, Swing desktop UI, file-backed persistence Advanced Parking System · NUML
  • Python Automation, data handling, REST API clients Advanced 3 × Michigan certs
  • Data structures & algorithms Complexity analysis and practical trade-offs Advanced NUML · 100% Michigan
  • Clean architecture & testing Layer boundaries, dependency inversion, widget tests Advanced CSOFT Systems

Delivery & Tooling

How the work actually ships
05
  • Git & GitHub Branch strategy, pull-request review, conflict resolution Production Google cert · daily
  • Figma → Flutter Design-token extraction, pixel-accurate handoff Production CSOFT Systems
  • Agile & Scrum delivery Sprint planning, estimation, standups, retrospectives Advanced Google cert · 97.5%
  • Vercel & static delivery CI previews, custom domains, edge cache headers Advanced This portfolio
  • Requirements & risk analysis Elicitation, prioritisation, secure-requirements practice Advanced 2 × CU Boulder certs
How it fits together

Five layers,
one clean boundary each

This is the shape every app I build takes. Widgets never talk to Postgres, the AI gateway never touches the UI, and the cache can go stale without breaking a screen. Hover the stack to pull the layers apart.

Presentation — Flutter widgets, 60fpsview
State — Riverpod providers & notifiersstate
Domain — repositories & use caseslogic
Data — Supabase, Postgres RLS, Groqremote
Cache — Hive encrypted offline storelocal
Live in Tripline

One itinerary request, end to end

What happens between a tap and a finished multi-day travel plan. The whole round trip lands in under a second on a good connection.

01 ~0 ms
Flutter captures intent

Destination, dates, budget ceiling and interests, validated on device.

02 ~40 ms
Context assembly

System contract, output schema and geolocation injected into the prompt.

03 ~600 ms
Groq Llama 3.3 70B

Streamed generation constrained to strict JSON, validated before use.

04 ~120 ms
Persist & render

Written to Postgres under RLS, then drawn as map markers and day cards.

Need one of these in your stack?

Every tier above is checkable — the code is on GitHub, the credentials verify on Coursera, and the commercial work is at CSOFT Systems.

See the work → Verify credentials
Hire Me →