Loader
Modelgraphs · recommendation · detection
Detecttraffic · evidence · investigation
Operateinfrastructure · deploy · on-call

I build systems.Then I try to findwhere they break.

Harshith Gangaraju

Enter the work

Cybersecurity · AI · Infrastructure

How I work

Don't just use it.

Understand it.

Break it.

Then build it better.

Position

I build software systems at the intersection of cybersecurity, AI and infrastructure.

Mostly I end up where those three meet. A detection pipeline is useless without the infrastructure to run it, and infrastructure nobody can operate is not finished. I like the whole path: the problem, the thing that stays up, and finding out what it does when someone leans on it.

Three overlapping fields: security, AI and infrastructure. Security and AI overlap in detection, AI and infrastructure in ML pipelines, security and infrastructure in operations. All three overlap in the whole path.SecurityAIInfrastructure

Right now I’m in the second year of a B.Tech, Computer Science (Cybersecurity) at Lendi Institute of Engineering & Technology, in Visakhapatnam. Class of 2029.

Selected work

01OperateCTO

Qyverix

A web studio I am technically responsible for: the product, the infrastructure it runs on, and the operations around it.

Read the case studyqyverix.in (opens in a new tab)

  1. Client brief
  2. Buildsite · CMS · commerce
  3. Deploycloud, environmentsDomains & mailDNS, webmail
  4. Accessauth · RBAC · secrets
  5. Operationslogging · backups · support

02DetectDesign & build

CYBER-OS

One detection and investigation system, not a verdict engine: it produces evidence a person can actually audit.

Read the case studygithub.com/hotaro6754/CYBER-OS (opens in a new tab)

  1. Network trafficunidirectional, read-onlyURL · Email · SMS · QRsubmitted for analysis
  2. Zeekconnection telemetryPlaywright sandboxSSRF-guarded
  3. RedpandaKafka-compatible busFeature extractionentropy, lexical, DOM
  4. DetectionXGBoost · Isolation Forest
  5. Evidence Fabricdetector version + input hash
  6. Caseentity graph · SOC view

03ModelDesign & build

Ideaspace

A campus innovation platform built around proof: ideas are cheap, so the system scores evidence instead.

Read the case studygithub.com/hotaro6754/Ideaspace (opens in a new tab)

  1. PeopleIdeasProjects
  2. Innovation graphthe relationships, not a feed
  3. Health scorecompleteness, 0–100ProofGitHub evidence, reviewed
  4. Points engineevery event logged with a reason
  5. Leaderboard · Archive

Practice

What I optimise for, in order. The size is the rank.

  1. 01CorrectnessIt does the thing it claims to do.
  2. 02SecurityIt still does that when someone is trying to break it.
  3. 03SimplicityThe smallest design that survives the real problem.
  4. 04MaintainabilityI can open it in six months without dread.
  5. 05UXSimple on the surface, even when it is not underneath.
  6. 06PerformanceFast, once it is right.
  7. 07ExperimentationLast, and only when the rest holds.
Sometimes five services and three agents should have been one good function.

Where I have had to mean it

  • CTOQyverixArchitecture, infrastructure, security, the 2 a.m. pages
  • FounderInfoShareStudent technology community
  • Marketing HeadAI ClubOutreach reaching 200+ students

Real systems don't care that you're still a student.

Currently

building

  • CYBER-OSCorrelating detections into one auditable case
  • Qyverix CloudThe infrastructure layer I wanted to use myself
  • This siteThe landscape is generated, not drawn

learning

  • Data structures & algorithms
  • Operating systems
  • NetworkingPast the point where the diagrams stop
  • Distributed systems

exploring

  • Offensive securityBreaking the assumptions everyone else made
  • Detection engineeringNoise into one useful signal
  • AI on real systemsCode, traffic and identity. Not chat.

Updated September 2026

Contact

If you are building somethingthat shouldn’t break,

let’s talk.
cto@qyverix.in

or, personally, harshith6754@gmail.com