Lil Task X

We turned your Product Manager into an AI, please don’t tell HR.

Comment

Inspiration

Product managers work through entire product lifecycle, from strategizing, planning the roadmap, creating requirements, managing sprints, coordinating with the team, creating and managing Jira, updating the stakeholders, and validating the GitHub repos. Most of this work is manual, repetitive, and scattered across tools.

They constantly fight against:

  • unclear stakeholder asks
  • budget and time constraints
  • inconsistent engineering effort
  • fragmented communication
  • undocumented work
  • missing context between Jira tasks and actual code
  • surprise mismatches (design says “blue,” repo says “green”)

We wanted to rethink PM work from the ground up and turn all this manual labor into something the PM doesn't have to worry about.

So we built Lil Task X, an intelligent PM system that takes constraints, people, tasks, and code, and turns them into strategy, execution, and delivery, all through AI.

What It Does

Lil Task X is a full-stack AI Product Manager that automates everything from idea to Jira to engineering testing to delivery and repo validation.

1. Product Strategy & Ideation

  • Takes stakeholder features and business goals
  • Analyzes competitors using SerpAPI
  • Performs market sizing and industry insights
  • Identifies scenarios (Base, Stretch, Budget-Constrained)
  • Produces a complete market-ready product description
  • Generates personas, risks, KPIs, and positioning

2. Requirements & Development Planning

The user provides:

  • Employee data
  • Tester data
  • Budget cap
  • Deadline
  • Product description

The model will:

  • Break features into stories and tasks
  • Build acceptance criteria
  • Prioritize tasks using RICE / MoSCoW logic
  • Split work across sprints based on timeline
  • Assign tasks based on skill and seniority
  • Ensure human-balanced work is being provided
  • Calculate cost and compare it against the budget

Then it returns:

  • a full sprint plan
  • role-based assignments
  • required budget vs available budget
  • options if constraints cannot be met

Example model outputs:

  • “This plan works as-is.”
  • “This plan works only if you increase budget by $14,200 or cut costs.”
  • “This plan works if we remove Feature #3 and extend timeline by one sprint.”

3. Jira Automation & Notifications

After the PM approves the plan:

  • Jira tasks are created automatically
  • Each engineer gets assigned a task based on your CSV
  • An exam is sent to engineers through SMTP about their tasks

4. GitHub Repository Cross-Checking

Lil Task X continuously monitors your repo to:

  • Match code commits to the Jira tasks
  • Examines commit messages such as:
    • “Done - Feature #1 Sprint #1”
  • Mark features as “Ready for Testing”
  • Assign Jira bug validation to testers
  • PMs are notified about new updates

Semantic code scanning includes:

  • Scanning the repository structure
  • Reading code comments
  • Extracting component names and logic
  • Comparing the code to each feature requirement

Example:
Stakeholder wants a blue UI, but repo uses a green theme. Lil Task X detects the mismatch, and then alerts the PM.

5. Delivery & Stakeholder Reports

Lil Task X generates:

  • Several PDF reports
  • Pie charts of budget and the progress
  • Jira + code alignment score
  • The probability of on-time delivery

This turns Lil Task X into a complete product delivery system.

How We Built It

Frontend

  • React + React Router
  • Supabase (secure storage)
  • Auth0 (authentication)
  • Feature forms, analytics dashboards, PDF viewer
  • CSV file upload UI

Backend

  • FastAPI
  • Python
  • Gemini (primary reasoning + planning model)
  • LangChain for RAG + structured explanation
  • SerpAPI for competitor/market intelligence
  • SMTP for sending notifications about the task
  • Jira REST API for live issue creation
  • ReportLab for generating the PDFs

Repository Intelligence

  • NVIDIA Nemotron-Nano-12B-v2-VL (NIM endpoint)
  • Developed for code understanding
  • Maps repo code to the requirements
  • Automatic mismatch detection
  • Always sync with Jira

Data

  • employee_catalog.csv
  • testers.csv
  • budget.csv
  • stakeholder_features.json
  • GitHub repo
  • Market search results

Everything flows through one orchestrated pipeline.

Challenges We Ran Into

  • Making AI reason without hallucinating
  • Getting data from competitor reliably
  • Ensuring a fair task distribution
  • Managing two AI models (Gemini + Nemotron)
  • Parsing commit messages to connect code to Jira
  • Repo-to-code semantic mapping without false positives
  • Enforcing strategy to development to delivery mindset in the model

Accomplishments We’re Proud Of

  • Created a real PM co-pilot and not just a chatbot
  • It connects everything from Stakeholders to Jira
  • It automates the full product lifecycle
  • Finds repo to Jira mismatches effectively
  • Spring planning becomes human friendly

What We Learned

  • AI can automate PM workflows, but only when there is structure
  • Multi-model systems require a lot of reasoning chains
  • Engineers benefit when tasks are clearer and more human
  • Great products come from alignment not just execution

What’s Next for Lil Task X

  • CI/CD integration (auto PR creation, automated code fixes)
  • Prioritized alerts in Slack/Teams when it truly matters
  • Design system checker where we compare UI designs to live code
  • Several natural-language queries
    • “How close are we to delivering Feature 4?”

Final vision:
A fully autonomous AI Product Manager that can take a feature request and deliver the finished product.

Built With

Share this project:

Updates