Headline

Building the Investigation Platform for Customer Support | Cohesity | CSE @ IIT Kanpur

About

We're building an Investogation Platform for B2B support teams. Modern support teams don't struggle because they lack tools. They struggle because customer context is scattered across Slack, Jira, Salesforce, Confluence, emails, logs, and countless internal systems. Every complex issue becomes a manual investigation, and valuable knowledge is often lost once it's resolved. We're building a platform where humans and AI investigate customer issues together. It continuously understands ongoing investigations, brings together the right context, recommends the next best actions, and captures institutional knowledge so every resolved issue makes the next one easier to solve. If you're building or leading an enterprise support organization, I'd love to connect.

Location

Bisauli

AI stack

antigravity

codex

cursor

Work experience

Founder & CTO at Stealth AI Startup, Jul 2026 - Present · 2 mos

Member of Technical Staff II at Cohesity, Sep 2024 - Jul 2026 · 1 yr 11 mos

Member of Technical Staff at Cohesity, Jul 2023 - Sep 2024 · 1 yr 3 mos, Pune District, Maharashtra, India

Member Of Technical Staff at DevRev, May 2022 - Jul 2022 · 3 mos, Bengaluru, Karnataka, India, Worked on scalable ingestion systems for enterprise communication workflows. Key contributions: • Developed an AWS-based data ingestion pipeline for scalable email processing workflows • Implemented robust SMTP, IMAP, and MIME parsing logic to reliably process heterogeneous email formats • Integrated external systems and APIs to support communication synchronization workflows

Research Intern at University of South Carolina, Jan 2021 - May 2022 · 1 yr 5 mos, Remote, Conducted research under Prof. Biplav Srivastava in explainable AI, adversarial robustness, and ML for healthcare. Research contributions: • Proposed generalized black-box input-perturbation algorithms for generating precise and robust visual explanations in object detection systems • Designed adversarial attack strategies exposing vulnerabilities in popular explanation methods for deep neural networks • Established benchmarks on the MS-COCO dataset for evaluating explanation robustness and fidelity • Open-sourced a state-of-the-art face-mask detection system with explainability features Publications resulting from this work were presented at CVPR, AAAI Workshop (W3PHIAI), and CODS-COMAD.

Education

Bachelor of Technology - BTech, Computer Science, Indian Institute of Technology, Kanpur

Public links

github: https://github.com/akshay-gupta123