4
Systems Built
Software Engineering Portfolio
Software Engineer Building AI-Powered Systems
Computer Vision, Geospatial Platforms, and Mobile Applications
Software engineer building data-driven systems combining machine learning, spatial analytics, and scalable backend services.
82
Automated Tests Passing (verified Jul 2026)
1.5K+
Labeled Samples (AutoEIT-STS)
Engineering Focus
Systems-first product engineering
AI-Enabled Systems
Designing end-to-end systems where ML models are integrated into reliable engineering workflows.
Backend and APIs
Building service layers, data pipelines, and deterministic APIs for production use-cases.
Spatial and Mobile Platforms
Developing geospatial analytics and offline-first mobile applications with clear system boundaries.
Featured Projects
Complex systems with implementation proof
HealingStone
3D Reconstruction System for Fragmented Artifacts
Problem
Reconstruct artifacts from partial 3D fragment scans.
Approach
Geometric feature extraction pipeline with embedding-based fragment matching and alignment.
Pipeline
Point Clouds -> Features -> Embeddings -> RANSAC + ICP -> Global Graph
Technical Output
- Reconstructed 3D assemblies from fragmented inputs
- Alignment quality metrics and similarity matrices
- Deterministic geometric refinement path
TerraHerb
Plant Disease Detection Platform
Problem
Classify plant species and diseases from leaf images.
Approach
Image ingestion and augmentation pipeline with CNN inference service and API delivery.
Pipeline
Image -> Augmentation -> CNN -> Classification -> API Response
Technical Output
- 24/24 automated tests passing; live API verified (Jul 2026)
- 88-class model training and evaluation workflow
- Inference-ready architecture for deployment
UDIE
Real-Time Urban Risk Intelligence Platform
Problem
Compute real-time risk across a city from streaming urban event signals.
Approach
Spatial event-sourcing backend with deterministic risk-field computation over a PostGIS grid.
Pipeline
Event Signals -> Ingestion -> Risk Compute -> PostGIS Grid -> API + Dashboard
Technical Output
- NestJS/PostGIS backend + Swift iOS integration
- Dynamic risk field modeling over event streams
- Real-time spatial analytics delivery pipeline
LifeTrack
Offline-First Mobile Health Analytics System
Problem
Track personal health data privately, fully offline, with no cloud dependency.
Approach
Modular Flutter app separating domain logic from Drift/SQLite persistence with reactive Riverpod state.
Pipeline
Entry -> Validators -> Drift (SQLite) -> Riverpod Providers -> Dashboard + Insights
Technical Output
- Flutter + Riverpod + SQLite/Drift architecture
- Offline-first operation with local state integrity
- Modular dashboard and health metric tracking
Verified Evidence
Measured results from real runs
Measured · Jul 2026
HealingStone — End-to-End Reconstruction Run
Full pipeline executed on the bundled 10-fragment sample corpus via the
healingstone-run CLI. All ten candidate pairs aligned successfully.
$ healingstone-run --data-dir data/sample/3d
Pipeline complete — reconstructed_model.ply (1,859 points)
metrics.json
aligned_pairs: 10
successful_alignments: 10
mean_icp_rmse: 0.0097
mean_chamfer_distance: 0.309
reconstruction_completeness: 0.749
graph: 10 nodes / 9 edges
Measured · Jul 2026
TerraHerb — Test Suite & Live API
Full pytest suite passing, and the FastAPI service verified live: health, class
catalog, and the /identify upload path including its low-confidence guardrail.
$ pytest tests/ -v
======== 24 passed in 37.72s ========
api 11 · classes 4 · inference 4 · model 5
$ curl localhost:8000/health
{"status": "healthy",
"service": "terraherb-ai-substrate",
"version": "0.2.0"}
$ curl localhost:8000/classes
{"count": 88, ...}
Measured · Jul 2026
LifeTrack — Automated Flutter Tests
Executed with flutter test: model serialization, health-value validators,
Riverpod providers, and a widget smoke test of the app shell all pass.
$ flutter test
✓ Medication model — BaseHealthEntry mapping
✓ Medication model — serialize / deserialize
✓ HealthValidator — blood pressure
✓ HealthValidator — heart rate
✓ rotatingDiseaseProvider — returns 5 diseases
✓ didYouKnowProvider — returns fact
✓ LibraryTab — renders diseases
✓ App starts (widget smoke)
8 executed tests, 8 passing
System Design Snapshots
Diagrammed architecture flow
HealingStone Architecture
Point Clouds
↓
FPFH Extraction
↓
Siamese Embedding
↓
RANSAC + ICP
↓
Global Reconstruction
UDIE Architecture
Event Signals
↓
Ingestion and Validation
↓
Spatial Risk Compute
↓
PostGIS Risk Grid
↓
API and iOS Visualization
TerraHerb Inference Flow
Image Input
↓
Preprocess + Augment
↓
CNN Inference
↓
Class + Confidence
↓
API Response
Experiments
Hypothesis-driven technical investigations
Descriptor Study
Compare FPFH descriptors against learned embeddings for fragment matching stability.
Model Variant Study
Evaluate EfficientNet vs MobileNet trade-offs for TerraHerb accuracy and inference cost.
Spatial Risk Study
Assess deterministic spatial risk field sensitivity under event burst and sparse-signal conditions.
Open Source Engineering
Algorithm implementation, architecture, and data pipelines
HealingStone
3D reconstruction pipeline emphasizing geometric matching and learned embeddings.
UDIE
Geospatial event intelligence engine with backend compute and mobile integration.
TerraHerb
CNN disease classification pipeline with dataset-driven evaluation.
Technical Skills
Full-stack engineering with AI capability
Languages
Python
C++
Dart
Backend & Systems
FastAPI
REST APIs
PostgreSQL
PostGIS
Docker
Linux
AI & Data
PyTorch
TensorFlow
OpenCV
Open3D
NumPy
CUDA
Platforms & Tools
Flutter
Git
NestJS
Geospatial Analytics
Mobile Systems
Experience & Education
Academic journey and engineering practice
2022 – Present
B.Tech — Computer Science & Engineering
B.Tech · India
Coursework in algorithms, data structures, machine learning, and computer vision. Independent research into 3D reconstruction, geospatial systems, and mobile-first architectures.
2024
Independent Research — 3D Fragment Reconstruction
Self-Directed · HealingStone Project
Designed and implemented a full geometric reconstruction pipeline using Open3D, FPFH descriptors, Siamese embeddings, RANSAC, and ICP for aligning fragmented artifact scans.
2024
Applied AI Project — Plant Disease Classification
Self-Directed · TerraHerb Project
Built an end-to-end plant disease detection platform achieving ~97% top-1 accuracy on PlantVillage dataset. Delivered a FastAPI inference service with full training pipeline.
2023 – 2024
Systems Engineering — Urban Risk & Mobile Health
Self-Directed · UDIE & LifeTrack Projects
Developed a real-time geospatial risk intelligence platform using NestJS and PostGIS, plus an offline-first Flutter health analytics app with Riverpod state management and local SQLite persistence.
Resume
Preview and download
- Applied AI engineering projects and outcomes
- Core technical stack and systems exposure
- Education and internship targeting focus
Contact
Open to ML and Computer Vision internship roles
Interest areas: computer vision systems, geospatial intelligence, and applied AI engineering.
Location: India | Collaboration: Remote and hybrid opportunities.