Personal portfolio / system blueprintAI · SOFTWARE · SYSTEMS

>initializing profile...

Mukesh G.

Applied AI. Verified systems.
Practical impact.

Computer Science undergraduate specializing in AI and ML. Building across computer vision, data-driven applications, and network security verification.

Explore my work

IDEA → IMPLEMENTATION → EVIDENCEFIG. 01 / NTS

COMPUTER SCIENCE AI & MACHINE LEARNINGGitHub SCROLL TO INSPECT ↓

01 /

Selected work

Ideas, implemented.

NETWORK SECURITYverification: 43/43

NetTwin

Adversarial network verification that checks what a firewall actually does against what its policy promises.

ACL policiesLinux nftablesTCP/UDPNetwork namespaces

Inspect case study

Verification report / August 2026

43/43 PASS

real-firewall test cases validated

Stated verification pipeline · isolated network namespaces

APPLIED AI / October 2025

AI Travel Planner

A trip that fits a student budget.

Gemini 2.5 FlashGoogle Maps APIPrompt Engineering

Documented inputs, APIs, and output · request sequence not specified

WEB APPLICATION / June 2025

Dynamic Weather

The forecast, wherever you are.

OpenWeatherMap APIsGeolocationGeocoding APIs

Documented components · request sequence not specified

Project results are reported from the resume. Open a case file for scope and evidence.

02 /

Tools, connected.

Context over checklists.

A stack means more when you can see what it built. Explore the technologies used together across my projects and research.

Explore a system

Hover, focus, or tap a node.
Lines indicate co-use in the selected work.

DEPENDENCY MAP10 NODES / CO-USE

SELECTED WORK Multimodal nutrition platform

View accessible skills list & project connections

Multimodal nutrition platform

Grounding DINO · SAM 2 · DINOv2 · FusionWeightNet · Flutter · Next.js · FastAPI · MongoDB · Google Cloud Run · Socket.io

Data Science Intern · Indian Statistical Institute, Kolkata

FiftyOne · Perceptual image hashing · Hamming distance

AI ML Intern · National Institute of Electronics and Information Technology

Logistic Regression · Random Forest · SVM · KNN

Data Analyst Intern · Andhra Pradesh State Skill Development Corporation

Python · Pandas · NumPy · Matplotlib · Seaborn

NetTwin

ACL policies · Linux nftables · TCP/UDP · Network namespaces

AI Travel Planner

Gemini 2.5 Flash · Google Maps API · Prompt Engineering

Dynamic Weather

OpenWeatherMap APIs · Geolocation · Geocoding APIs

Complete skill inventory

Languages

Python · SQL · Java · Dart · TypeScript · JavaScript

AI / ML & data

Scikit-learn · NumPy · Pandas · Matplotlib · Seaborn · Google Gemini API

Computer vision

Grounding DINO · SAM 2 · DINOv2 · FiftyOne · Pillow · ImageHash

Web & applications

HTML · CSS · React · Next.js · Flutter · Node.js · REST APIs

Databases & AI tools

MongoDB Atlas · Mongoose · Generative AI APIs · Amazon Q · Codex · ChatGPT

Delivery

Git · GitHub · Streamlit Cloud · Vercel · Prompt Engineering

Coursework

Computer Networks · Database Management Systems · Object Oriented Programming · Operating Systems

03 /

The commit history.

Research & experience

eb15c19

Summer Research Fellow

Indian Academy of Sciences × IIIT Hyderabad

June 2026 – July 2026

Engineered a multimodal AI nutrition platform for multi-dish food detection, segmentation, classification, and portion-weight estimation.
Built and deployed a full-stack system with offline meal synchronization, real-time updates, and local-to-cloud AI failover.
Combined medical profiles, allergies, nutrient targets, and meal intake to generate health warnings, portion coaching, nutrition scores, and context-aware AI dietary recommendations.

Grounding DINOSAM 2DINOv2FusionWeightNetFlutterNext.jsFastAPIMongoDBGoogle Cloud RunSocket.io

View supporting work

02a2341

Data Science Intern

Indian Statistical Institute, Kolkata

August 2025 – October 2025

Studied image deduplication and visual similarity with Voxel51’s FiftyOne and perceptual image hashing.
Evaluated hash-based similarity on 51 real-world images, identifying duplicate and near-duplicate samples with similarity scores up to 100%.
Contributed to experimental evaluation of FiftyOne for visual data analysis and similarity modeling.

FiftyOnePerceptual image hashingHamming distance

View supporting work

0017a39

AI ML Intern

National Institute of Electronics and Information Technology

June 2025 – July 2025

Analyzed AQI classification using historical pollutant data from Indian cities.
Compared supervised models using accuracy, precision, recall, F1-score, and confusion matrices. Random Forest achieved 82% accuracy.

Logistic RegressionRandom ForestSVMKNN

View supporting work

efeac59

Data Analyst Intern

Andhra Pradesh State Skill Development Corporation

May 2025 – July 2025

Studied the relationship between smartphone specifications and battery-life performance across device categories.
Cleaned, aggregated, and compared battery-performance data using Pandas and NumPy.
Created correlation heatmaps, histograms, box plots, scatter plots, bar charts, and pie charts with Matplotlib and Seaborn.

PythonPandasNumPyMatplotlibSeaborn

View supporting work

Expand a record to inspect the work. Hashes are visual record IDs.

04 /

Beyond the implementation.

A little context

I’m a Computer Science undergraduate specializing in AI and ML. My work connects applied research with working software: a multimodal nutrition platform at IIIT Hyderabad, image similarity analysis at the Indian Statistical Institute, and adversarial firewall verification with NetTwin. I’m interested in real-world AI systems, and in understanding how they behave beyond the model.

Currently / in progress

Currently researching Adaptive-Horizon World-Model Planning. Open to projects, internships, and collaborations.

2026 – 2027 · Pursuing