Akash Shahriar

The dtac app is the mobile application of a leading telecommunications company in Thailand. The app is designed to give customers easy and convenient access to their services, including managing accounts, checking balance and usage, purchasing packages and promotions, and much more.
Here are some of the key features of the app:

• Account management: The app allows users to manage their dtac accounts, including updating personal information, viewing account details, and checking balance and usage.

• Package and promotion purchase: Customers can use the app to purchase various packages and promotions, including data, voice, and messaging plans.

• Payment management: The app allows customers to make payments and top up their accounts using various payment methods, including credit cards, debit cards, and e-wallets.

• Reward points: The app also offers a reward points program that allows customers to earn points for various activities, such as purchasing packages, paying bills, and referring friends. These points can be redeemed for different rewards, such as free data, voice, and messaging services.

• Store locator: The app can locate nearby dtac stores and service centers.

An Android-based video conferencing app used by several clients of Karel Telecommunication, Istanbul. I was tasked with updating their old codes with the latest dependencies and native libraries and fixing bugs. I have worked closely with RTP packets, MediaCodec, JNI, Opus, and Speex. This was an exciting project as the codes were 3-4 years old, which made me study RTPs and how different encoded video formats are transferred through UDP/TCP.

A video editing Android app that includes features like adding music to video, slow motion, fast motion, changing canvas size, Ai effects, and trimming. I have worked on the implementation of most of the app features, optimized the video engine with the Android native decoder, and implemented an optimized video reel generation system.

A React Native-based mobile platform for both iOS and Android where RV owners can find a technician for services while technicians can promote their services.

I worked on this project as a full-stack developer in building the following:
• Back end with Express.js and Node.js
• Admin panel with React
• Customer website with React
• Customer app with React Native
• Technician app with React Native

I have maintained the codebase, fixed multiple issues, and implemented new features. I deployed the back end and websites on the GCP and set up third-party libraries like GCP, AWS, MongoDB, Atlas, Twilio, and Mailgun. In addition, I also configured different environments for different build variants–development, staging, and production.

An Android photo editor app with more than ten million users ranked in the top ten in the photo editing app category. I led three team members in implementing new features and refurbishing the app's version 4 as I designed the architecture, memory management, and optimization.

Logoshop is a logo maker app with a modern landing page where users can find many logo templates, select one, and edit it or create a new one from scratch. I have worked on this app's editing panel and output system, designing the editing panel's framework and doing all the touch and experience-related work. I have implemented the output generation system and was in the R&D team adding 3D text using canvas.

An android-based photo background eraser that can erase the background from a picture using image segmentation. It is a highly optimized and fast app that can efficiently segment the background from a picture. I have designed the architecture and system of the cut-out feature. This feature has a great editing panel where users can remove the background and create stickers. Users can then set any background image and also add stickers or text.

The Blur background photo editor app has many one-tap editing tools to take your picture art experience to the next level. Blur background photo editor is a free image editor with plenty of fun stickers and DSLR camera effects. It offers incredible filters and brings the most trending features, such as spiral and portrait maker. Now you can apply neon borders and wings to your images and make them look stylish in a cartoon style.

This point-of-sale system provides an excellent replacement for a cash register and tracks sales and inventory in real-time; it can manage your sales and increase them with good management. Being a suitable replacement for the cash counter, it enables the mobility to sell anywhere, and in the way the customers want to buy.

The products allow you to instantly share your contact info, social media, files, videos, and more with just a tap. The best part, the other person doesn't need an app or our products to receive your info.

Kaaz is a pioneer job engagement platform designed to connect employees and employers for casual and informal employment in a peer-to-peer fashion. Although the app is available globally, the launch is piloted in Kenya and Bangladesh. We help people find job opportunities or expert casual workers whenever and wherever they need them; you can simultaneously be a job seeker or a skill seeker. Download the app and explore limitless earning possibilities by monetizing your skills through simple steps.

BCS Prostuti is an educational app designed to aid aspirants in their preparation for the BCS preliminary exam, as well as other government job exams, including banks and primary school teachers. It is a comprehensive platform for learners with 45000+ questions, where exclusive questions with detailed explanations are continuously added.

RS Sjoliv is a map-based app designed for boaters in Norway, providing local information about ports, marinas, and other boats near them. The app features an emergency button that sends the user's location to emergency services, and a messaging service called Sjovenn for social interaction and advice among boaters. The app also allows users to filter their search based on their preferences.

In this project I:
• Led a cross-platform mobile application development project using Flutter and Dart, with a team of five developers and two designers, following industry-leading software development best practices.
• Utilized GraphQL to communicate with AWS AppSync and MQTT clients, ensuring efficient data retrieval and communication between the app and the devices.
• Implemented efficient state management using ValueNotifier and ChangeNotifier, which offers more flexibility than third-party state management libraries like Getx, Reiverpod, and Bloc, ensuring a seamless and uninterrupted user experience.
• Implemented a device compatibility solution that allowed the Flutter app to adapt dynamically to different device types, platforms, screen orientations, and sizes, ensuring a seamless user experience across all platforms.

Binge is an all-in-one streaming service that offers its users a vast array of entertainment options. The platform provides over 3000 original web series, movies, dramas, award-winning TV shows, documentaries, kids' content, and live TV channels.

The project faced several challenges during its development, and I had a key role in:
• Implementing and optimizing adaptive streaming and playback of various media sources using ExoPlayer, resulting in efficient buffering and seamless playback.
• Optimizing the code architecture following best practices and clean architecture.
• Migrating the app code from Java to Kotlin.
• Contributing to the development of new features and functionalities based on user feedback and market trends.

An AI-based restaurant ad generator and AI-powered marketing SaaS tool. With this tool, users can post ads with the help of AI and automatically customize the audience engagement with the help of segmented data in the warehouse.

PhenomenAI is a future-proof GenAI video creation ecosystem to add intelligence to your creativity, connections, community, and workflow.

I led full-stack development on this project, building both the front end (Next.js, React, Android) and the back end (NestJS, Python). I architected the system with JWT-based authentication, implemented WebSockets for a seamless real-time editing and generation panel, and integrated an SQS queue for asynchronous video processing. I also worked on video processing pipelines and used LangChain to build a prompt-enhancement agent. On the infrastructure side, I designed and deployed the AWS architecture using Amazon ECS, CloudFront, load balancing, and DocumentDB.

A Flutter-based marketplace and investment platform where people post different marketing and investment videos (Reels). In short, it is a TikTok for investors and businesses. I have built the landing page, website admin portal, and Android and iOS apps. I have developed the full cycle from scratch to deployment.

OVERVIEW
An all-in-one AI lifestyle companion designed to enhance user productivity and well-being through personalized coaching, mood tracking, and daily organization.

KEY CONTRIBUTIONS
Integrated LLM Capabilities: Engineered the core AI chat functionality using OpenAI's API to provide empathetic, coherent responses for emotional support and personalized coaching.

ADVANCED TRACKING SYSTEMS
Developed a modular architecture for tracking mood and menstrual cycles, translating physiological signals into actionable health insights.

AR AND UTILITY FEATURES
Implemented an "AR Share" feature for augmented reality social interactions and a "Yummy" meal planner that integrates with grocery APIs to suggest recipes based on available ingredients.

STYLIST ENGINE
Built an interactive fashion recommendation engine that utilizes user preference quizzes and brand databases to suggest tailored outfits.

EMERGENCY SUPPORT
Engineered a high-priority "SOS" help feature allowing users to trigger emotional crisis support with specific keyword triggers.

TECHNICAL STACK
Flutter, FastAPI, OpenAI API, Augmented Reality (ARCore/ARKit), and RESTful API integrations.

I serve as the lead architect for "Music Near Me," a real-time event discovery and studio booking platform for the music industry. Originally built on FlutterFlow and Firebase, I spearheaded the complete migration to a custom full-stack Flutter mobile application with a Supabase back end and a Next.js administrative portal.

The migration was driven by the need for complex relational data handling and specialized geospatial queries that exceeded the capabilities of the original low-code setup. I architected the new system to leverage Supabase’s PostgreSQL capabilities for real-time venue availability and event synchronization. On the front end, I developed a highly responsive Flutter app for iOS and Android, ensuring seamless map integration and low-latency performance. On the business side, I built a robust admin dashboard with Next.js to enable venue owners to manage high-concurrency booking schedules. This refactor significantly improved app performance, data integrity, and scalability across multiple regions.

I designed and built AIDeskHub, a multi-tenant AI voice receptionist SaaS that answers business calls 24/7. To solve the friction of setup, I engineered an onboarding pipeline where users simply paste their website URL; the back-end multi-page crawler extracts business data and uses Gemini to auto-generate a structured agent knowledge base. I architected the core telephony system on Retell AI, managing speaker-turn transcriptions, sentiment analysis, and lead classification. To ensure reliability under heavy production loads, I enforced strict multi-tenant isolation at the database level using PostgreSQL and built custom Stripe metered billing. I also implemented robust webhook handlers that safely process event payloads asynchronously to keep data in sync. I secured the back end with over 790 automated tests across 14 distinct API modules, ensuring zero-regression releases.

I architected and shipped GangIt, a high-performance B2B SaaS platform for DTF print shops. I designed a customized, browser-based canvas editor using Fabric.js that operates in real physical inches rather than raw pixels to eliminate scaling errors. On the back end, I engineered a high-resolution rendering pipeline built on NestJS and Sharp that composites overhanging graphics and auto-rotates EXIF photos to output production-ready 300 DPI print sheets (up to 45.75″ × 240″), rendering massive 987-megapixel files under strict memory and timeout envelopes on Google Cloud Run. Additionally, I built the entire marketplace and billing infrastructure, including multi-tenant custom pricing, automated US sales tax calculation via Anrok, and secure seller payout logic using Stripe Connect Express with a 7-day settlement delay to protect against chargebacks.

I developed BoxOfDox, a HIPAA-compliant SaaS that automates the medical record review process for personal injury attorneys. To process hundreds of pages of scanned PDFs, I built an RAG pipeline utilizing recursive chunking and a FAISS vector store generated dynamically per case folder to guarantee data privacy. To maximize recall of scattered clinical facts, I configured a wide retrieval range of k=40 and designed a section-by-section draft generation flow using GPT-4. I engineered a secure clinician-review workflow that stores the raw AI draft and the licensed physician's signed edits separately, ensuring full legal provenance for court-ready reports. Additionally, I built a secure audit logging layer tracking every system and user action with IP and user-agent details, and implemented a tiered pricing system with Stripe integrations.

I co-designed and shipped Fitness Room Concierge, a comprehensive maintenance-as-a-service platform consisting of a Flutter mobile app, a Next.js admin dashboard (51 screens), and a FastAPI back end. On the mobile side, I built a Flutter client supporting offline-first Isar database caching and integrated Zebra thermal-printer engines via Bluetooth and network to output physical asset QR labels directly on the gym floor. I modeled a complex six-role permission hierarchy and a state-machine that routes work orders based on whether a machine's metadata indicates it is under manufacturer warranty. I engineered a load-balanced preventative maintenance scheduler and wired nine real-time push notifications using Firebase Cloud Messaging.

I designed and delivered ParkSight, a peer-to-peer parking marketplace where hosts monetize empty spaces, and drivers book guaranteed parking. To guarantee transactional integrity, I engineered a server-side pricing engine that signs quotes into short-lived, cryptographically signed tokens, preventing client-side price tampering. The database was architected to only create a booking record after a successful Stripe webhook callback, completely eliminating unpaid-state errors. Additionally, I built an Express Event Parking module with atomic slot inventory to handle stadium-concurrency rushes without double-booking. The host financial engine manages a secure payout state machine with automated fund reversal if a Stripe Connect withdrawal fails. On the client side, I built the map-based discovery app in Flutter, utilizing marker clustering for heavy listing densities.

I architected Cogology, a bilingual (English/Spanish), HIPAA-compliant mental health assessment platform. To move away from volatile, unconstrained chat loops, I designed the patient screening assistant on LangGraph. This structured the AI dialogue as a stateful graph with deterministic nodes and transitions, ensuring the conversation predictably guides patients through validated clinical instruments (such as GAD-7, PHQ-9, and ISI) along a safe clinical path. On the back end, I built a rule-driven routing engine that analyzes a short mobile pre-screen to automatically assign indicated clinical questionnaires to the patient's dashboard before their clinic appointment. I also delivered a 19-route Next.js clinician portal that structures manual session notes into a unified five-step clinical format and automatically aggregates longitudinal diagnostic scores.

I engineered MediMind, an offline-first medication reminder app that translates photos of handwritten paper prescriptions into complete dosing schedules. Recognizing that LLMs naturally output plausible-looking but dangerous hallucinations, I implemented a strict three-stage validation pipeline: every extracted drug name must match, validate, and enrich against official clinical medicine databases before setting a reminder. The reading pipeline uses a self-hosted MedGemma model for medical shorthand, with a Google Gemini Vision API fallback for degraded images. The client-side database was built on Isar to allow patients to access their reminder schedules, take-dose confirmations, and history logs without an internet connection. To streamline development, I also extracted and tested a reusable internal Flutter framework (easy_flutter) secured by unit tests.