PROJECTS
Codebase Janitor & Documentation Bot - Python, FastAPI, GitHub WebHooks
OBJECTIVE
Automating technical documentation and repository structure maintenance through AI-driven
diff analysis.
PROBLEM STATEMENT
Manual README updates often lag behind code changes, leading to 'documentation rot' and
making repositories difficult to navigate for contributors.
SOLUTION
A GitHub App that uses the Groq Meta Model to analyze git diffs in real-time. It automatically
generates README updates, project summaries.
KEY FUNCTIONS & MODULES
- verify_request() : Verifies if the request received is sent by GitHub Webhook or anything
else by using HMAC encoding.
- generate_summary() : Sends the PR diff data to the GROQ AI model to generate a
summary of the provided diff.
- receive_data() : An async function that parses the request sent by the GitHub webhook
and filters out relevant information.
- construct_data() : Constructs the separate text of the summary generated by the Janitor
bot without modifying the prior text.
- send_data() : Sends the constructed data back to the GitHub webhook via a PATCH
request which updates the existing body of the PR and adds the generated summary to the
PR comment with a headline, “Generated by Janitor Bot :”
TECHNICAL HIGHLIGHTS
Leverages asynchronous programming to ensure that documentation generation does not block
the development workflow, providing near-instant feedback on repo updates.
API-Based Voice Classifier - Python, FastAPI, RandomForest, Scikit-Learn
OBJECTIVE
Providing a high-performance, locally-deployable API for speech and audio recognition across
multiple input formats.
PROBLEM STATEMENT
Cloud-based voice classification services often carry high latency, subscription costs, and privacy
concerns for sensitive audio data.
SOLUTION
An optimized API service that utilizes a Random Forest classifier for local inference. It handles
diverse inputs including direct file uploads and base64 encoded strings for seamless integration
into various apps.
KEY FUNCTIONS & MODULES
- verify_api_key() : Verifies if the request contains the correct API key or not.
- extract_features() : Extracts features like MFCC, ZCR, Centroid, RMS, Pitch and
converts them into one vector for the RandomForest model.
- get_audio_from_request() : Checks for the audio format sent in the request. It checks
for formats like base64, url, or a file (.mp3 or .wav).
- predict_from_file() : Loads the model into the API, converts the extracted features
vector into an array and sends it to the model to check for probability and predictions and
returns 0 for AI and 1 for Human based on the predictions.
TECHNICAL HIGHLIGHTS
Designed with privacy in mind, ensuring all audio processing remains on the local hardware
without external data transmission.
Data Filtration And Visualization Tool - ReactJS, Python, FastAPI, NumPy, Pandas,
ChartJS
OBJECTIVE
Enabling non-technical users to perform complex data analysis and visual storytelling through a
minimalist web interface.
PROBLEM STATEMENT
Raw CSV/Excel datasets require manual cleaning and expert knowledge of spreadsheets to
derive meaningful visual insights.
SOLUTION
A full-stack application that leverages Pandas for high-speed backend data manipulation and
React for a clean, 'flat' UI that renders interactive charts based on filtered results.
KEY FUNCTIONS & MODULES
- FileUpload : This React module allows the user to simply drag and drop their
dataset file or manually choose it.
- FilterPanel : This React module is the section where the user can apply filters on
their dataset and also allows them to see their dataset with only 100 rows as
preview after applied filters.
- VisualizationPanel : This module is the section where the user can visualize their
dataset in different chart forms like histograms, pie-charts, scatter-graphs etc.
- dataframe_info() : This function returns the metadata about the dataset such as
columns and their data type etc.
- filter_data() : This function is where filters are applied to the uploaded dataset or
dataframe and returns the preview for it.
TECHNICAL HIGHLIGHTS
Implements a minimalist design philosophy with a steel blue and pale gray color palette,
ensuring a focused and modern user experience.