Maria Aqdas@maria_aqdas
6
Events attended
3
Submissions made
Pakistan
2+ years of experience
About me
I’m a BS Data Science student from Pakistan with a strong interest in AI and real-world applications. I have experience working with Python and data analysis tools like NumPy and Pandas, and I’m actively building projects to strengthen my skills. I enjoy solving problems, learning quickly, and collaborating with teams to create impactful solutions during hackathons.
🤓 Latest Submissions
Dukaan Dost: AI Stock Forecast for Kiryana Stores
# Inspiration & Problem Neighborhood grocery (kiryana) shops in emerging markets operate on thin margins, high cash-flow constraints, and informal supplier delivery cadences. Most shopkeepers restock based on gut feel or basic static averages, resulting in frequent stockouts of high-velocity staples like milk, tea, and flour, or excess dead capital tied up in slow-moving inventory. # What Dukaan Dost Does Dukaan Dost turns daily sales logs into actionable inventory intelligence: - **Statistical Demand Forecasting:** Models seasonal weekday demand variations and level-smoothed trends across products over a rolling 7-day horizon. - **Dynamic Stockout Prediction:** Evaluates supplier visit cycles and delivery lead times to compute precise restock triggers, unit counts, and estimated restock costs. - **Rigorous Verification & Simulation:** Features a transparent 28-day historical replay simulator that proves stockout reduction versus traditional baseline ordering rules without data leakage. - **Accessible, Multilingual AI Ordering:** Uses AMD Token Factory open models to convert mathematical forecasts into human explanations and supplier WhatsApp restock messages in both English and Roman Urdu, backed by a strict anti-hallucination guardrail that blocks any draft containing unverified numbers. # Architecture & Tech Stack - **Frontend & Interface:** Lightweight, responsive vanilla web architecture with custom canvas-based data visualizations, printable distributor restock slips, and bilingual localization toggles. - **Forecasting Engine:** Deterministic level-smoothing and weekday factor algorithm executed client-side. - **AI & Acceleration:** Serverless API integration powered by AMD Token Factory running open open-weights LLMs for contextual explanation and verified supplier message generation.
18 Oct 2026
Upkeep Console Codebase Maintenance Copilot
Problem Statement: In fast-paced software development, technical debt is always visible, but engineering teams rarely have the dedicated bandwidth to fix it until an outage occurs, a pull request review slows down, or a new engineer questions legacy functions. Stale dependencies with known vulnerabilities, dead code branches, unhandled promise rejections, and missing retry timeouts accumulate quietly over time. Traditional manual audits take 14+ hours per repository and result in ad-hoc, inconsistent patches that frequently introduce unexpected regressions. Our Solution: Upkeep Console is an autonomous codebase maintenance copilot built on IBM Bob 2.0 that transforms reactive maintenance into a scheduled, auditable, and automated workflow. Instead of waiting for technical debt to cause system failure, Upkeep points Bob 2.0 at live service repositories (demonstrated on checkout-service) to proactively detect, classify, and resolve maintenance debt before it becomes critical. Target Users & Workflow: Target users include backend engineers, DevOps leads, and repository maintainers responsible for enterprise microservices. The maintenance workflow operates through a structured 5-gate pipeline: 1. Detected: Bob scans the entire repository to uncover stale dependencies, dead branches, and untested execution paths. 2. Triaged: Every identified issue receives an objective risk and effort score; nothing is modified prior to classification. 3. Bob Fixing: Items marked safe are dispatched to parallel Bob subagents, each resolving issues on its own isolated branch. 4. Verified: Every automated fix is re-run against the full test suite before being marked complete. 5. Shipped: Verified fixes ship with clear changelogs, while high-risk structural decisions are cleanly handed off to maintainers. Creativity, Uniqueness & Impact: Upkeep Console combines repository-level context with an interactive, in-browser static analyzer that checks for exposed secrets, unhandled promises, and console leftovers. In live demonstration benchmarks on checkout-service, Upkeep resolved 7 out of 9 debt items autonomously, raised test coverage from 61% to 88%, eliminated all known CVEs and dead code branches, and passed 41/41 regression tests in just 9 minutes—saving over 13.8 hours of manual engineering effort.
27 Sep 2026
NeuralVoice AI Assistant
NeuralVoice AI Assistant is an innovative, real-time voice-interactive web application built to bridge the gap between human speech and advanced artificial intelligence. Designed as an accessible, high-performance solution, it allows users to interact seamlessly using natural voice commands, receiving instant, accurate answers and audio feedback without manual typing. Key Features & Architecture: Speech-to-Text & Recognition: Integrated with SpeechRecognition and AssemblyAI to accurately capture, transcribe, and process user audio input. Google Gemini AI Engine: Powered by the modern google-genai SDK and the high-speed Gemini flash models, providing intelligent, contextual, and lightning-fast textual responses. Text-to-Speech (TTS): Features automated voice synthesis using pyttsx3 to read responses aloud back to the user, creating a fully hands-free conversational loop. Live Telemetry & Dashboard: Built on Streamlit with a custom dark-purple UI, offering real-time tracking of inference speeds, model status, and a dynamic chat conversation stream. Secure Cloud Deployment: Fully configured for secure deployment on Streamlit Cloud using environment secrets for API key management. Technologies Stack: Python, Streamlit, Google Gemini API, AssemblyAI, SpeechRecognition, Pyttsx3, and Dotenv.
30 Sep 2026
