I'm a software engineer from Ahmedabad, India. I joined MindArmor.ai, US-based cybersecurity startup, as an early engineer and grew into Co-founder & CPO, taking ownership of the platform's technical and product direction. Built the system from the ground up across multi-agent AI, LangGraph, RAG, evaluation, AI reliability, backend architecture, AWS infrastructure, and full-stack product development. Designs AI systems around orchestration, validation, failure handling, and observability rather than treating model output as the product itself. Works across architecture and product — translating ambiguous customer problems into systems that can be deployed, operated, and used in real-world environments.
I joined MindArmor as an engineer and now lead the product and technical direction. I own the platform end-to-end, working across AI architecture, backend systems, cloud infrastructure, security, and the product experience.
Designed AI-driven incident simulations mapped to an organization's real infrastructure, assets, controls and response teams, replacing static tabletop exercises with runs that test how decisions actually get made under pressure.
Developed AI pipeline: a LangGraph orchestrator hands work to specialist sub-agents, and a critic/validator step checks their output before it reaches users, to catch hallucinated security-training content.
Fixed silent failure where fabricated results were being returned as successful responses. Added proper failure states, retries and structured logging.
Integrated enterprise SSO and identity-provider authentication with role-scoped access, so participants, facilitators and auditors each reach only what their role requires.
Set up separate dev and prod environments with PR-based promotion, and took the Cognito auth flow through a full security review.
Conducted discovery and demo calls with clients and turn what they ask for into shipped features.
Jan – Jul 2025
Astraum Digital
Software Engineer · Ahmedabad
Worked on healthcare products focused on connecting patients with home-care services and attendants, contributing across product engineering, application architecture, and deployment
Built and shipped patient and attendant-facing workflows for healthcare service discovery, booking, scheduling, and service management.
Designed reusable React/Tailwind components and application flows that supported multiple healthcare use cases while keeping the product consistent and maintainable.
Integrated frontend applications with REST APIs, handling authentication, data flows, validation, and asynchronous application states.
Worked directly with design, backend and leadership to scope features.
Contributed to the AWS deployment and infrastructure, improving the path from development to production and supporting reliable application delivery.
Took ownership of features from requirements and technical design through implementation, integration, testing, and deployment.
education
2021 – 2026
K.S. School of Business Management and Information Technology
Integrated M.Sc. in Information Technology · Ahmedabad
projects
Self-optimizing outreach engine
A system that does B2B outreach on its own: it researches a target company, finds the right decision-makers,
and writes a personalised first message. Research runs as fixed pipelines, while LangGraph agents handle
strategy and recover when a step fails. Every message is checked against schemas and cited sources, protected
against prompt injection from scraped pages, and needs a human's approval before it's sent. It learns from past
campaigns, but any change to its strategy also needs human sign-off.
A web app for job seekers who want to reach companies directly instead of only applying on job portals.
Users connect their own email account with OAuth, and ColdReach helps them send personalised cold emails
to companies and track what they've sent, with checks that keep emails landing in the inbox rather than spam.
Built with React, Node.js and MongoDB; 100+ people signed up in the first week.
Alpine FIBC manufactures FIBC bulk bags, the large woven bags used to ship things like grain, cement and
chemicals. Off-the-shelf ERPs didn't fit how their factory works, so I built a custom one that brings their
orders, production and inventory into a single system their team uses day to day.
A payment management system for a business group that runs several separate firms. Before, each firm handled
its payments on its own. AstraumPay puts all of them in one place: vendor payments going out to suppliers,
and internal payments between the group's own firms, so the owners can see and manage money across every
firm from a single dashboard.
A home and interiors platform where AI turns traditional architectural plans into realistic views of the
actual room. Instead of trying to read a 2D drawing, a customer can see what the finished space will look
like before anything is built.
Content aggregation platform
A web platform that pulls content from multiple sources into one place and lets users search across all of it.
It has its own login system and paid subscriptions through a payment gateway. Built with React on the
frontend and Django on the backend.
open source
GitHub
Contributed to 12+ open-source projects and opened 80+ issues.
See github.com/DevGajjar28.
AnalySim is an open platform for researchers to share and analyse large datasets. Before writing the
proposal I ran it locally and audited the codebase. I built a working researcher/admin dashboard in
the project's own stack (Angular, Bootstrap, Plotly, D3) and walked through it with the mentor.
While doing that I found four access-control vulnerabilities in the production system and disclosed
them to the maintainers with reproduction steps.
The proposal covered integrating that dashboard, a settings page, a fixed project-creation flow,
a better CSV data browser, and patching all four vulnerabilities with integration tests.
A plan to take end-to-end test coverage of the OpenELIS Global React frontend from about 30% to 80%
with Cypress, and run the suite on every pull request through GitHub Actions. I worked in the
codebase and submitted pull requests to the project during the application period.
An early-warning system for Aadhaar enrolment and update data, built in a team of three.
We cleaned and merged about 4.9 million records across three datasets (including fixing 18+ misspelled
state names), then built ARIMA forecasts, IQR and MAD-based anomaly detection, and a five-signal risk
score down to pincode level. Everything sits in a 12-tab Streamlit dashboard with a Gemini assistant
that only answers questions about the dashboard.
A REST API that takes a voice clip and says whether it's AI-generated or a real person, across
English, Hindi, Tamil, Telugu and Malayalam. It extracts pitch, MFCC, spectral centroid, energy and
zero-crossing features, classifies with a logistic regression model, and falls back to rules for
borderline cases. Returns a label, a confidence score and a short explanation. Deployed with Docker
on Hugging Face Spaces.
2025
SSIP Hackathon: ShikshaSetu
A monitoring system for the 68 municipal schools under Bhavnagar Municipal Corporation, so education
officers can follow attendance, teacher reports and school activity without visiting every school.
Includes a school ranking, a parent portal and alerts for low attendance. Team of five.
React, Tailwind, Node/Express, Python, PostgreSQL
2025
Odoo hackathon: FarmSure
A marketplace connecting natural farmers directly with buyers. Farmers are verified by an admin and
get an on-chain certificate; buyers scan a QR code on the product to check where it came from, and
earn eco points for their purchases. Team of four.