Kark Robo

ML enthusiast and published researcher

Rajkot 0+ yrs exp 86 · Excellent

About

ML enthusiast and published researcher with a B.Tech in Computer Engineering (CGPA 9.46/10, GATE DA 2026 AIR 7007). Experienced in deep learning, federated learning, and end-to-end data engineering — from CNN training under privacy constraints to cloud pipelines and production API deployment. Proficient in PyTorch, Python, SQL, Docker, FastAPI, and AWS. Looking to build real things at the intersection of ML and meaningful problems.

Skills & Expertise (26)

PyTorch Advanced

9.0/10

Years Exp

CNNs Advanced

9.0/10

Years Exp

Federated Learning Advanced

9.0/10

Years Exp

Explainable AI Advanced

9.0/10

Years Exp

Python Advanced

9.0/10

Years Exp

Pandas Advanced

8.5/10

Years Exp

NumPy Advanced

8.5/10

Years Exp

scikit-learn Advanced

8.5/10

Years Exp

TensorFlow Advanced

8.5/10

Years Exp

Supervised Learning Advanced

8.5/10

Years Exp

Unsupervised Learning Advanced

8.5/10

Years Exp

Prompt Engineering Intermediate

8.0/10

Years Exp

Predictive Modelling Intermediate

8.0/10

Years Exp

Statistical Analysis Intermediate

8.0/10

Years Exp

AWS S3 Intermediate

7.5/10

Years Exp

SQL Intermediate

7.5/10

Years Exp

Matplotlib Intermediate

7.0/10

Years Exp

Docker Intermediate

7.0/10

Years Exp

REST APIs Intermediate

7.0/10

Years Exp

Git Intermediate

7.0/10

Years Exp

Snowflake Intermediate

6.0/10

Years Exp

Power BI Intermediate

6.0/10

Years Exp

JavaScript Intermediate

6.0/10

Years Exp

FastAPI Intermediate

6.0/10

Years Exp

Postgresql Intermediate

6.0/10

Years Exp

Vue.js Intermediate

6.0/10

Years Exp

Work Experience

AI Strategy & Business Intelligence Intern

CSRBOX × AICTE (IBM SkillsBuild)

Mar 2026 - Apr 2026

Developed KisanSeva, an AI-powered agricultural assistant built with n8n and Supabase, delivered under IBM’s SkillsBuild programme (ID: 2026AICSIB0870).

Research Intern

Adroid Connectz

Jan 2026 - Mar 2026

Trained a CNN on BloodMNIST using differentially private SGD (DP-SGD) within a federated setup, achieving >93% diagnostic accuracy under strict privacy constraints (ε < 30). Quantified the privacy–accuracy–explainability trade-off: reducing ε from 66.46 → 28.91 produced a 3.8% accuracy drop and 11.2% reduction in LRP explanation consistency across federated clients. Formulated the three-way interaction between differential privacy, model accuracy, and XAI as a constrained multi-objective optimisation problem, directly applicable to production ML systems. Independently authored two publication-ready research papers, covering literature review, experimental design, formal problem formulation, and academic writing.