Kishaan Gidda

Cybersecurity & AI/ML · Ottawa

I build the thing,
then I try to break it.

Hello, I'm Kishaan. I'm a computer science co-op student at Carleton University in the AI/ML stream, and right now I'm on a cloud security product team at TrendAI, building agent tooling that triages security alerts. I love both halves of this work: shipping the system, then hunting for the way in.

Get in touch [ View work ]

Currently · Product Team, AVTD at TrendAI

01 · Selected work

Work

Three things I've built or won: an optimal-control solver I couldn't stop working on, an AI triage system with real stakes, and a capture-the-flag weekend that ended first place.

A circuit drawn as two hairline track edges with a dotted geometric centreline, and over it the minimum-lap-time racing line (wide at entry, tight at the apex, wide at exit), with the apexes marked.

FastAPI · CasADi/IPOPT · React · Three.js

Racing Line App

Draw a track on flat ground or real-world terrain and the solver returns the minimum-lap-time racing line. Then orbit it, or drive the optimal lap from a driver's-eye camera in 3D. Two solvers sit behind it: a fast curvature heuristic, and a periodic optimal-control problem over 360 collocation nodes solved with IPOPT through CasADi. 15K+ lines and 590+ tests later it also trains its own drivers in the browser, evolving neural-net policies with CMA-ES across parallel Web Workers.

A diagram of CareRouter's routing: an intake is scored by a Gemini-based severity assessment, then routed down one of four bands: low to self-serve resources, moderate to a community clinic, elevated to matched urgent care, and crisis directly to a crisis line without queueing.

FastAPI · Gemini 2.5 Flash · 2026

CareRouter

An AI mental-health triage system that gets people to the right support fast. I was the sole backend developer on a four-person team: a rule-constrained Gemini pass scores severity, urgency and issue type from a six-question intake, a second pass filters real facilities on cost, transport and language, and crisis cases route deterministically to 9-1-1 or 9-8-8 without ever touching the model. Safety-first by construction, not by policy.

[ Source ↗ ]

CyberSci results listing: Ottawa Region, Carleton University, Team "The Off-By-Ones"
Competition results · Ottawa region

Capture the flag · 1st of 13 teams

CyberSci Ottawa Regional

A room full of people trying to break the same systems I spend my weeks defending, which is my favourite way to spend a weekend. The Off-By-Ones finished first out of thirteen teams at the Ottawa regional.

[ CyberSci ↗ ]

02 · Experience

Experience

  1. Current

    TrendAI

    Product Team, AVTD

    • Built a read-only SOC alert-triage agent on the Claude Agent SDK that chains multi-step tool calls across six custom MCP tools to triage alerts and correlate cloud-asset CVEs
    • Fixed a two-region production incident where Azure zonal-capacity failures left zombie VMs marked healthy and stranded 30+ scans, then killed the failure class for good with idempotent teardown and launch guards
    • Shipped AWS region eu-south-2 end to end for agentless vulnerability scanning and VPC Flow Logs (CloudFormation, pipelines, tests), then codified the 15-task runbook into a reusable Claude Code skill
    • Expanded weekly BlackDuck and Fortify SCA/SAST scans from 22 repos to all 52 with GitHub Actions, then remediated 270+ critical and high findings across 23 repos, driving the critical backlog to near zero
    • Cut a live nine-service disaster-recovery simulation from 10 hours to 5, orchestrating AI subagents through three end-to-end dry runs to clear every CI failure
    • [ Claude Agent SDK ]
    • [ MCP ]
    • [ AWS ]
    • [ Azure ]
    • [ CloudFormation ]
    • [ Vulnerability management ]
  2. May – Aug 2025

    Hydro Ottawa

    Cybersecurity Analyst Intern

    • Built a third-party app-risk pipeline that scrapes Google Workspace Marketplace listings and scores each app's terms of service and security policies with Gemini's REST API
    • Wrote four incident-response playbooks, 20 to 30 pages each, covering AI data leaks, web exploits, ransomware and insider threats, then validated every one of them end to end in test environments
    • Deployed AWS GuardDuty with Terraform to secure S3, built a Lambda pipeline that auto-quarantines malicious files with SNS alerts, and migrated 20+ PGP secrets into Secrets Manager
    • Emulated Living-off-the-Land attacks by chaining 10 to 25 MITRE ATT&CK abilities in Caldera, which is where I learned how quiet a real intrusion looks
    • Tested DNS connections with a PowerShell script driving Cloudflare API calls
    • [ AWS ]
    • [ Terraform ]
    • [ MITRE ATT&CK ]
    • [ Gemini API ]
    • [ IAM ]
    • [ PowerShell ]
  3. Feb – Aug 2023

    Ciena

    Software Developer Intern

    • Automated daily test-result reports with a cron-scheduled Python tool, writing SQL against Oracle and emailing formatted tables to every deploying team. It is still running today, which I am still quietly proud of
    • Built the Angular and TypeScript front end of an internal IAM permission-assignment tool
    • Learned DevOps and Agile in practice, with Jira and Git
    • [ Python ]
    • [ SQL ]
    • [ Angular ]
    • [ TypeScript ]
    • [ Oracle ]
    • [ Linux ]

03 · About

About

Education
Carleton University · Computer Science Co-op, AI/ML stream 3.6 GPA · Dean's Honour List
Focus
Agentic AI · optimal control · offensive security · capture-the-flag
Languages
English · French, DELF B2
Off keyboard
Soccer, playing and watching Liverpool · the gym, most days · reading on where AI is actually heading

Toolkit

Marked · what I use most

Languages

  • [ Python ]
  • [ TypeScript ]
  • [ Go ]
  • [ C++ ]
  • [ Java ]
  • [ SQL ]

AI & optimisation

  • [ Claude Agent SDK ]
  • [ MCP ]
  • [ Gemini API ]
  • [ ONNX ]
  • [ CMA-ES ]
  • [ IPOPT ]

Cloud & security

  • [ AWS ]
  • [ MITRE ATT&CK ]
  • [ Linux ]
  • [ Terraform ]
  • [ Docker ]
  • [ PowerShell ]

Web & tooling

  • [ FastAPI ]
  • [ React/Next.js ]
  • [ GitHub Actions ]
  • [ Git ]

04 · Contact

Have something to build,
or something to break?

I'm looking for my next co-op or internship in AI/ML and security engineering. If you're building agents, hardening cloud, or both, I would genuinely love to hear about it. My inbox is open and I answer everything.

kishaangidda@gmail.com

[ GitHub ↗ ] [ LinkedIn ↗ ] [ Blog ]