Free tools Windows power users keep installed
One-click scans. No signup required.
Over the past year, Sufiyan Khan, a developer from Satara, India, says he has built five projects that put AI into practical web workflows: document search, lead follow-up, code review, report writing, and bookings. His clearest technical takeaway is that building a RAG app is only part of the job; checking whether its answers are useful and grounded takes deliberate evaluation.
What the projects do
| Project | Input and task | How the output is checked or used | Evidence described |
|---|---|---|---|
| Groundwork | Uploaded documents and user questions | Combines keyword and semantic search, reranks results, and generates answers that point to source material. | Khan reports Ragas evaluation scores; no dataset or run configuration is given. |
| LeadAI | New sales leads | Drafts and emails follow-ups, then logs leads in Google Sheets through an n8n workflow. | No measured conversion or business-impact results are reported. |
| AI CodeReview Pro | Code changes | Uses a group of agents to review changes; risky areas such as authentication and payments are sent to a person rather than automatically approved. | The human handoff is a described safety design choice, not a measured reduction in defects. |
| NotesToReport | Raw notes | Creates a report with citations and blocks it when claims do not match the notes, using a stated faithfulness threshold. | Khan reports a threshold of 0.85; its behavior is not independently validated. |
| Aurelia Retreat | Booking inquiries | Pairs a booking website with an AI assistant and a lead dashboard. | No evaluation results are stated. |
Groundwork: retrieval is not the same as answer quality
Groundwork is the most detailed example of the year’s technical work. Khan describes a retrieval flow that combines keyword matching with semantic search, reranks retrieved material, and produces answers linked to their sources. That design aims to make document-based responses easier to trace back to the material a user supplied.
As an Amazon Associate I earn from qualifying purchases.
He also reports a deployment problem: the search index disappeared while the app was using free hosting, so he moved index storage to Qdrant Cloud. That account illustrates a practical distinction in RAG projects: a promising retrieval and generation flow still depends on durable storage for its index.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →For evaluation, Khan says he used Ragas and reports a faithfulness score of 1.00 and an answer-relevancy score of 0.71. These are his project-specific, self-reported results, not a general benchmark or independently replicated test. The surfaced article does not specify the evaluation dataset, metric configuration, or run conditions, so the figures should not be compared directly with scores from other apps. Within that limit, the gap he reports points to a useful iteration question: an answer may stay faithful to retrieved material while still not fully addressing what the user asked.
#1 Best Overall
Automation with a person still in the loop
LeadAI handles the first follow-up steps
LeadAI targets delays after a new lead arrives. Khan says the n8n-based workflow drafts and emails a follow-up, then records the lead in Google Sheets. The described workflow automates message preparation and record-keeping; the article does not report conversion rates or show that it improves sales outcomes.
AI CodeReview Pro escalates higher-risk changes
The code-review project assigns review work to a group of agents, but does not treat every change as safe to approve automatically. Khan says changes involving sensitive areas such as authentication and payments go to a person. That is a clear boundary for automation: AI can assist with review while a human retains responsibility for decisions the builder considers risky. No defect-rate measurement is provided.
Grounding generated reports and booking assistance
NotesToReport constrains claims to the notes
NotesToReport turns raw notes into a cited report. Khan says it blocks the report if its claims do not match the notes, applying a faithfulness cutoff of 0.85. That number is the project’s stated threshold, not a universal standard for report quality; the article does not provide independent validation of how reliably the gate works.
Aurelia Retreat adds an assistant to a booking site
Aurelia Retreat combines a booking website with an AI assistant and a lead dashboard. It shows another use of AI in a conventional full-stack experience: the assistant sits alongside the booking flow and the tools for handling prospective guests. The article does not provide usage or quality measurements for the site.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What the year’s work suggests for builders
- Make outputs inspectable. Groundwork’s source-linked answers and NotesToReport’s citations give users a way to relate generated text to its input material.
- Evaluate specific qualities, not a vague idea of “good.” Groundwork’s reported faithfulness and answer-relevancy scores are distinct measures, and the lower reported relevancy score gives Khan a concrete area to work on.
- Choose where automation must stop. LeadAI automates follow-up tasks, while AI CodeReview Pro routes certain sensitive changes to a person. The appropriate handoff depends on the consequences of an incorrect action.
- Plan for the dependencies behind the feature. Groundwork’s lost index on free hosting shows why persistence and deployment choices matter as much as the retrieval design.
Khan’s post is a first-person project retrospective and portfolio presentation, not an independent product comparison. He points readers to his portfolio for project demos and code and says he is taking on freelance AI automation and full-stack work; those details are his own account.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
