Akash Kundu

Fine-tuned a CNN to 85%+ accuracy across 7 eye diseases, built an offline voice AI assistant with sub-50ms inference, and cut model training time 70% by troubleshooting GPU pipelines from scratch. Final-year Computer Science student who applies that same research-build-debug-deploy discipline across NLP, computer vision, and agentic workflows. Looking for an AI internship where that hands-on habit can plug into a team's actual roadmap.

AI Research and Social Innovation Intern - Purposive AI - Kolkata (Hybrid, Part-time)

(2026-07)

• Architecting Bengali voice-AI pipelines in Linux CLI environments, shipping production-ready workflows into Kotha-Khata's WhatsApp deployment for rural Self-Help Groups.
• Reduced onboarding friction for non-technical Self-Help Group users, as measured by fewer support escalations during rollout, by automating bookkeeping and government-scheme accessibility workflows.

AI & ML Intern (Supported by Google for Developers) - AICTE & EduSkills - Virtual

(2025-10 - 2025-12)

• Completed 100% of a 10-week ML curriculum, building and debugging NLP and neural-network experiments in scikit-learn and TensorFlow via CLI.
• Improved model accuracy across 4+ assessments, as measured by assessment scoring, by engineering features and automating dataset-experimentation pipelines.

B.Tech - Computer Science & Engineering - Techno India University (2027-09)