Evidence catalog / recruiter view

Applied systems, reviewed in the open.

I build software across applied AI, data systems, Android, and security-conscious product engineering.

This static route separates shipped, documented, in-progress, and early work. Every claim is bounded by a reviewed repository, publication, or approved documented source.

01 / Selected

Featured Systems

The strongest current evidence across data engineering, portfolio architecture, and documented UAV research.

ShippedRepository verified

Enterprise Data Processing Platform

A local Python data-processing system with a CLI and authenticated FastAPI interface for validating synthetic order CSVs, storing accepted records, explaining rejections, and producing repeatable reports.

Role / contribution

Independent portfolio project by Anup Kumar Dagala, built with AI assistance and synthetic data.

Problem

Make incoming order-file quality, duplicate handling, persistence, and operational reporting explicit and reproducible in a local system.

Implementation

A typed Python package shares validation, processing, storage, and reporting services across argparse and FastAPI interfaces, with locked dependencies, automated tests, CI, and a non-root container definition.

Technologies

Python 3.13
FastAPI
Pydantic
Pandas
SQLite
pytest
GitHub Actions

Security / reliability decisions

Shared-key authentication on business routes
Parameterized SQL and transactional integrity
Request, row, concurrency, rate, and retained-record limits
Generated storage paths and spreadsheet-formula neutralization
Documented threat model and CI security checks

Limitations

A local learning demonstration using synthetic data, not a production SaaS or company deployment.
One shared key, one process, one SQLite writer, and no user roles or distributed rate limiter.
The recorded benchmark is ingestion-only on one laptop and is not a throughput target or SLA.
Public deployment would require production identity, TLS, retention, backup, monitoring, and operational controls.

In progressRepository verified

Anup Portfolio, World Lab, and Atlas

An in-progress Next.js portfolio exploring a semantic recruiter route, a progressively enhanced 3D journey, and a deterministic local evidence guide.

Role / contribution

Designed and implemented as Anup Dagala’s portfolio, with explicit architecture and security review at each milestone.

Problem

Present a varied engineering portfolio without making the core evidence dependent on WebGL, animation, or an opaque retrieval system.

Implementation

Server-rendered semantic content remains primary; browser-only 3D and Atlas interaction are isolated at small client boundaries with deterministic fallbacks and exact registries.

Technologies

Next.js
TypeScript
React
React Three Fiber
Three.js
CSS Modules

Security / reliability decisions

No user-controlled asset URLs, imports, camera coordinates, or selectors
No Markdown or unsafe HTML rendering in Atlas
Bounded same-origin API input with no persistence or prompt logging
Application-owned citation and action registries

Limitations

World Lab remains experimental and is not the production homepage.
Atlas is deterministic matching, not an LLM or complete RAG backend.
The audited Atlas checkpoint is local to this branch and may be ahead of the public default branch.
The site is not deployed to anupdagala.com.

DocumentedDocumented source

SWARM — Smart Wireless Aerial Reconnaissance Mechanism

A documented three-UAV architecture for coordinated missions in GPS-denied indoor environments.

Role / contribution

Documented system architecture and research work.

Problem

Coordinate sensing and mission activity among multiple aerial vehicles where GPS is unavailable indoors.

Implementation

A three-UAV architecture combining local perception, obstacle-aware navigation, distributed sensing, and telemetry coordination.

Technologies

UAV systems
Computer vision
Telemetry
Distributed sensing

Security / reliability decisions

Public wording excludes confidential implementation detail and unsupported deployment claims.

Limitations

No commercial deployment or public-safety adoption is claimed.
No production-readiness or unreviewed performance metrics are claimed.
Completed hardware validation is not asserted without reviewed public evidence.

02 / Supporting

Projects and Research

Documented systems with explicit evidence boundaries and limitations.

DocumentedDocumented source

Drone-Based Crowd Monitoring and Control System

A documented aerial crowd-analysis system using computer-vision workflows for density and movement estimation.

Role / contribution

Documented applied computer-vision system work.

Problem

Use aerial visual information to help operators understand crowd density and movement during safety-sensitive situations.

Implementation

A documented computer-vision workflow for aerial density and movement estimation with an operator-oriented situational-awareness goal.

Technologies

Computer vision
UAV systems
Image processing

Security / reliability decisions

Biometric-identification claims are explicitly excluded from the public description.

Limitations

This is documented work, not a deployed public-safety system.
No public performance metrics or operational-impact claims are included.

DocumentedDocumented source

Embedded FSM Cyber-Physical Security Framework

An ESP32-oriented cyber-physical security design using traffic metadata, cumulative threat scoring, and a finite-state policy engine.

Role / contribution

Documented cyber-physical security design.

Problem

Coordinate escalating network and physical responses on a constrained embedded platform using explainable state transitions.

Implementation

An ESP32-oriented finite-state policy design driven by network-traffic metadata and cumulative threat scoring.

Technologies

ESP32
Finite-state machines
Network metadata
Embedded systems

Security / reliability decisions

The design separates observed traffic metadata, scoring, policy state, and physical response actions.

Limitations

No measured attacker-dwell-time reduction is claimed.
The entry documents a design rather than a production deployment.

In progressDocumented source

Android-Based UAV Mission Platform

An in-progress Kotlin Android interface for mission planning, telemetry visualization, operational status, and live UAV monitoring.

Role / contribution

In-progress Android mission-interface work.

Problem

Give UAV operators a mobile interface for mission planning, telemetry, status, and live monitoring.

Implementation

A Kotlin Android interface with modular communication workflows for exchanging mission data.

Technologies

Kotlin
Android
Telemetry
Mission-data exchange

Security / reliability decisions

Public content distinguishes implemented direction from planned capability.

Limitations

The platform is in progress and has no reviewed public source link yet.
Planned capabilities are not represented as complete or field-deployed.

03 / Practice

Professional Experience

User-approved summaries that omit confidential company implementation details.

DocumentedDocumented source

Bhashini AI Solutions — 2025 Internship

Worked on Python-based OCR and data-processing workflows, with integration across Java wrappers and Kotlin Android components.

Role / contribution

Software engineering intern, 2025.

Problem

Support reliable OCR and data-processing workflows across backend and Android integration boundaries.

Implementation

Python extraction, preprocessing, and validation work integrated through Java wrappers and Kotlin Android components, with testing and documentation support.

Technologies

Python
OCR
Image preprocessing
Java
Kotlin
Android
REST APIs

Security / reliability decisions

Company data, private architecture, customers, and confidential implementation details are excluded.

Limitations

This entry is based on reviewed user-approved documentation rather than a public repository.
It does not imply ownership of the complete company product.

DocumentedDocumented source

Bhashini AI Solutions — 2024 Internship

Supported ERPNext workflow configuration and Python OCR integration for SQL-backed data workflows.

Role / contribution

Software engineering intern, 2024.

Problem

Support ERP workflows and OCR-backed extraction within SQL-oriented backend processes.

Implementation

ERPNext configuration support combined with Python OCR integration, SQL-backed data extraction, debugging, testing, and workflow documentation.

Technologies

ERPNext
Python
SQL
OCR

Security / reliability decisions

Company data, internal architecture, customers, and private screenshots are excluded.

Limitations

This entry is based on reviewed user-approved documentation rather than a public repository.
It does not imply ownership of the complete company product.

04 / Research

Publication

Peer-reviewed work with a fixed public publication source.

ShippedPublication verified

Transcriptome sequencing and differential expression analysis in bacterial wilt tolerant and susceptible clones of tomato (Solanum lycopersicum L.) following Ralstonia solanacearum infection

Co-authored published research involving transcriptome and differential-expression analysis in tomato clones following Ralstonia solanacearum infection.

Role / contribution

Co-author; approved contribution wording includes data preprocessing, statistical analysis, visualization, software, and formal analysis.

Problem

Study differential gene-expression responses in bacterial-wilt tolerant and susceptible tomato clones following infection.

Implementation

Published transcriptome research with differential-expression analysis, data preprocessing, statistical analysis, visualization, software, and formal analysis contributions.

Technologies

Data preprocessing
Statistical analysis
Visualization
Transcriptome analysis

Limitations

The portfolio does not assign experimental responsibilities beyond the approved contribution record.
The publication link is evidence metadata; Atlas does not fetch the article at runtime.

05 / Archive

Early Work

Learning projects retained with their implementation gaps and unsupported claims made explicit.

Early workRepository verified

Emotion-Aware Food Recommender

An early Streamlit image-processing prototype that captures a visitor-initiated webcam image, preprocesses it, applies a pretrained FER model, and maps the detected emotion to food suggestions.

Role / contribution

Early image-processing prototype; no model-training or measured quality claim is made here.

Problem

Explore whether a simple preprocessing path can make webcam emotion input easier to inspect before mapping results to recommendations.

Implementation

A Streamlit app decodes a captured image, applies OpenCV preprocessing, runs the same pretrained FER detector on original and processed images, and displays both outputs.

Technologies

Python
Streamlit
OpenCV
FER
TensorFlow
Pillow

Security / reliability decisions

The camera input is explicitly visitor initiated within the Streamlit interface.

Limitations

No recorded comparison supports improved-accuracy or illumination-robustness claims.
Dependencies use open version ranges and no lockfile is present.
No tests, CI, license, FER2013 attribution detail, or webcam privacy documentation are present.
One caught detector exception is displayed directly to the visitor.

Early workRepository verified

Tomato Leaf Disease CNN

An early tomato leaf-classification repository whose audited commit contains a Flask inference wrapper, Dockerfile, README, and interface images.

Role / contribution

Early computer-vision learning repository; model performance and end-to-end reproducibility are not claimed.

Problem

Explore a web-facing image-classification workflow for identifying tomato leaf disease categories.

Implementation

The audited source contains a Flask wrapper around an imported prediction function, API-key middleware for one route, and a Waitress launch path; required supporting implementation is missing.

Technologies

Python
Flask
Waitress
Docker

Security / reliability decisions

The archival summary does not repeat unverified model-quality or deployment claims.

Limitations

Training, prediction, model, dependency, template, and static files referenced by the code or README are absent.
Accuracy, robustness, reliability, real-time behavior, and dataset provenance are not reproducibly evidenced.
CORS is enabled globally, one prediction route is unauthenticated, and request-size limits are not visible.
No tests, CI, release, or license are present.

Early workRepository verified

Chai Time Note Taker

An early React interface experiment with in-memory note creation, search, colour selection, and reset controls.

Role / contribution

Early frontend learning work retained as an honest archive entry.

Problem

Experiment with a lightweight visual workspace for composing and finding short notes.

Implementation

React function components hold note and modal state in memory and render a simple board; persistence, secure authentication, and a functional music integration are absent.

Technologies

React
JavaScript
CSS

Security / reliability decisions

The public description explicitly treats the login as presentation-only, not authentication.

Limitations

No package.json or lockfile is present, and source imports expect a missing components directory.
The login is presentation-only and provides no identity or access control.
The Spotify API client has no token configuration and the search flow is unfinished.
The default Create React App README and stale generated test do not document actual behavior.

07 / Method

Engineering Principles

Make claims traceable to reviewed evidence.
Keep core content usable without optional graphics or client JavaScript.
Treat user input and generated output as untrusted at every boundary.
Prefer explicit limits, deterministic behavior, and honest limitations.

08 / Sources

Source Links

Only reviewed public destinations are linked from this route.

Anup Dagala on GitHub