I build AI systems and product-grade software that have to survive beyond demos: reliable interfaces, measurable outcomes, maintainable code paths, and engineering decisions grounded in production reality.
I work at the intersection of applied AI, product engineering, and security-aware systems. My focus is not only model output, but the complete path from idea to usable workflow: data handling, interface clarity, deployment constraints, reliability, and user-facing proof.
| AI/ML Systems | Computer vision, applied ML workflows, agentic-system research, model-backed product features |
| Product Engineering | React, TypeScript, Flask, Django, APIs, dashboards, portfolio-grade UX, deployment-ready builds |
| Security & Networks | CCNP/CCNA foundation, defensive analysis, network operations exposure, authorized testing context |
| Research Translation | Turning papers and experiments into systems with measurable behavior, usable flows, and clear documentation |
| 3 Research / publication tracks |
14+ Verified certifications |
CCNP / CCNA Networking credentials |
~40% Attack-surface reduction from authorized NMU penetration test |
| 4+ Portfolio-grade applications |
15 Open upstream PRs under active triage snapshot |
Meta / Sophos / DPIIT Cross-domain exposure |
Active OSS Docs, tooling, and systems contributions |
- Production AI/ML systems that move from experiment to usable workflow with clear operator value.
- Full-stack platforms where frontend quality, backend reliability, and product clarity are engineered together.
- Applied research projects translated into deployable software instead of being left as isolated papers.
- Security-conscious systems shaped by networking, defensive analysis, and real operational constraints.
Professional portfolio platform built with React, TypeScript, and Vite to present technical case studies, research, services, and engineering proof in polished product form.
Engineering signal: frontend execution, component discipline, visual systems thinking, public technical positioning, and production-grade presentation for technical work.
React TypeScript Vite UI Engineering Portfolio Systems
Interactive India-focused air-quality platform with city comparison, AQI storytelling, health context, forecast handling, and live/fallback API resilience.
Engineering signal: public-data product thinking, resilient API integration, forecast UX, fallback behavior, and user interpretation beyond a basic dashboard.
Data Products Forecast UX API Integration Frontend Systems
Plant Disease Recognition System
Flask and PyTorch application for crop disease detection with offline-friendly workflow, Marathi-ready UX, generated reports, weather context, and farmer-oriented recommendations.
Engineering signal: applied ML, computer vision, deployment constraints, report generation, regional usability, and practical decision support for non-trivial end users.
Flask PyTorch Computer Vision Applied ML Offline-Friendly UX
| Contribution | What moved |
|---|---|
| github/docs#44778 | Clarified workflow_dispatch environment input behavior |
| Panniantong/Agent-Reach#387 | Documented Windows Twitter cookie limitation |
| nexus-substrate/nexus-eval-atbench#32 | Removed broken lint script |
| mozilla-mobile/firefox-ios#34168 | Consolidated setup guidance and follow-up context |
- 2026-07-08: Rebuilt the profile README into a sharper proof-first engineering profile; updated project framing, proof signals, research wording, and recent-log visibility. Daily log
- 2026-06-29: Opened docs issues in trending repos: browser-use/video-use#92 and altic-dev/FluidVoice#471; rechecked active PRs and GitHub notification signals. Daily log
- 2026-06-23:
headroomlabs-ai/headroom#1084closed as superseded after project-header safety work reachedmain; branch contribution history preserved. Daily log
Multi-Angle Industrial Inspection Fusion
Viewpoint-invariant defect detection system with 95.3% accuracy, 0.991 AUC, and 31.4 FPS edge inference.
Paper: IJVRA2603948
Thermal + Depth Fusion for Predictive Maintenance
Multimodal fault anticipation pipeline designed to detect maintenance risk up to 72 hours before failure.
Paper: IJVRA2604277
FONTA: Failure Ontology for LLM Agents
Ontology-driven analysis of autonomous agent failures across 1,200 trials, focused on reasoning breakdowns, failure patterns, and evaluation structure for agentic systems.
| Organization | Role | Domain |
|---|---|---|
| Meta | Jr. Network Analyst Intern | Network operations, infrastructure exposure, production connectivity context |
| Sophos | Jr. Security Analyst Intern | Security analysis, defensive workflows, threat-aware thinking |
| DPIIT, Government of India | Tech Support Intern | Public-sector systems support, technical troubleshooting, operations discipline |
Building systems where:
- AI usefulness has to survive real product constraints.
- Frontend polish and model capability are held to the same standard.
- Reliability, security thinking, and measurable outcomes matter as much as raw feature count.
- Research claims are connected to artifacts, workflows, and verifiable engineering proof.
If you are building product-grade AI workflows, applied ML tools, or engineering-heavy software with real users, I am interested in that problem space.
- Open contribution notes live in
contributions/. - PR snapshots live in
pull-requests/. - Repository hygiene uses
CONTRIBUTING.mdand profile repo standards.