TaskFlow.ai
The perfect all in one for your company
Created on 22nd June 2025
•
TaskFlow.ai
The perfect all in one for your company
The problem TaskFlow.ai solves
💼 TaskFlow.ai — Intelligent Project & Workforce Management SaaS
🚀 Built in 36 hours to solve the chaos of project management, employee overload, and inefficient meetings.
🧠 The Problem
Modern IT companies rely on fragmented tools to manage people, projects, and productivity:
- HR systems to manage employees
- Separate project management tools (Jira, Trello)
- Time trackers (Toggl, Clockify)
- Manual spreadsheets to assign work
- Video calls with no action items
- Overloaded teams and unclear task ownership
This leads to:
- 💥 Work overload and burnout
- ❓ Tasks lost in conversation
- 🔁 Repetitive, manual work assignment
- ⚙️ Low efficiency and delayed projects
✅ Our Solution: TaskFlow.ai
TaskFlow.ai is a full-stack SaaS platform that:
- Centralizes company, employee, and project management
- Uses AI to transcribe meetings and generate task summaries
- Auto-assigns jobs based on availability, skill, and project priority
- Tracks work hours and load to prevent burnout
- Makes every meeting actionable — no wasted talk time
🎯 Who is it for?
- IT service companies
- Product teams
- Freelancing agencies
- Remote teams with recurring project work
🔧 Key Features
| Category | Features |
|---|---|
| 👥 Company Management | Register company, add employees, manage roles |
| 📁 Project & Task | Create projects, define priorities and deadlines, assign tasks |
| 🕒 Time Tracking | Start/stop work, manage time slots, prevent overload |
| 🎥 Meetings | Built-in video calls (Jitsi), auto recording |
| 🧠 AI Automation | Transcribe meetings, summarize key points, auto-create task list |
| ⚙️ Smart Assignment | Match tasks to best-fit employees by skill and availability |
| 🧑💻 Dashboards | Separate views for Admins, Employees, and Project Managers |
🧪 Demo Workflow
- ✅ Login/Register as a company admin
- 👥 Add employees
- 📁 Create a project with priority and required skills
- 🧑💻 Start a meeting → it records and transcribes the discussion
- 🧠 AI summarizes the meeting and creates tasks
- ⚙️ System auto-assigns tasks to the right team members
- 🕒 Employees track their time and see upcoming work
- 📊 Admin views team workload and task progress
💡 Built With
- 🧩 Frontend: React + TailwindCSS
- 🔧 Backend: FastAPI + PostgreSQL
- 🧠 AI/ML: Whisper (Transcription) + GPT-4 (Summary & Task generation)
- 🎥 Video Conferencing: Jitsi Meet integration
- 🔒 Auth: JWT-based secure login
- ⚙️ DevOps: Docker (optional for deploy)
🏁 Why It Matters
TaskFlow.ai makes meetings productive, manages teams smartly, and helps companies hit deadlines without breaking their people.
No more juggling 5 tools. No more overburdened teams.
Just clear meetings, smart assignments, and balanced workloads.
Challenges we ran into
🚧 Challenges We Ran Into
1. ⏱️ Time Constraints with a Multi-Module System
We aimed to build a complete SaaS platform in just 36 hours — with modules for employee management, project tracking, AI transcription, smart task assignment, and real-time communication. Prioritizing what to build and what to cut was crucial.
2. 🤖 Whisper Transcription Setup
Setting up OpenAI Whisper locally proved difficult due to limited hardware resources. Transcription on CPU (no GPU available) was slow, and processing long meeting recordings within time constraints was a bottleneck.
3. 🧠 GPT-Based Task Extraction
Generating accurate and actionable tasks from raw meeting transcripts using GPT-4 required carefully crafted prompts. Mapping tasks to team members based on skill and availability had to be simplified due to time pressure.
4. 🖥️ Hardware & Resource Constraints
We worked on personal laptops and VMs with limited RAM and no GPUs. Running AI models (Whisper), Jibri (for recording), and PostgreSQL concurrently was resource-intensive. Some features had to be downscaled or mocked to avoid crashes and lag.
5. 🔌 Frontend & Backend Integration Bottleneck
Our frontend developer was limited to building UI only. This meant our full-stack member had to handle API integration, increasing their workload and introducing slight delays in completing the frontend flow.
6. 🎥 Jitsi + Jibri Integration
While Jitsi Meet integration via iframe was relatively straightforward, setting up Jibri for auto-recording introduced major challenges:
- High hardware demand (Jibri alone can consume 2–4 GB RAM).
- Lack of GPU slowed video encoding.
- Configuring audio loopback (using
snd-aloop
) in a VM environment was tricky. - Docker container setup was not trivial and consumed valuable hours.
7. ⚖️ Smart Task Assignment Logic
Designing a logic that assigns tasks based on skill, availability, and project priority was complex under time constraints. We opted for a basic rule-based matching system with hardcoded skills and weights for the demo.
8. 🔄 Service Synchronization & Pipeline Orchestration
Moving data between modules — from Jitsi → audio → Whisper → GPT → backend → frontend — required careful coordination. Managing intermediate data, delays, and ensuring the right API trigger at the right time took multiple iterations.
9. 🧪 Testing Under Time Pressure
We focused on testing key flows like company registration, project creation, meeting → transcription → task flow. However, edge-case testing and performance checks had to be postponed to maintain delivery speed.
10. 🧠 Team Efficiency vs Feature Ambition
Not all teammates could contribute at full capacity due to complexity in unfamiliar areas. We had to downscale features like real-time transcription, advanced scheduling algorithms, and fancy dashboards to ship a working MVP.
Tracks Applied (3)
Authenticate with Civic Auth
Civic Technologies
Best use of Gemini API
Major League Hacking
Best Use of MongoDB Atlas
Major League Hacking