Complete Engineering Portfolio (22 Systems)

Production Platforms, AI Agents & Case Studies

Every project below is rendered directly in physical static HTML for crawlers and humans alike. Each case study details real software shipped, architecture choices, what I learned, and what I now know, paired with embedded video demos.

2026 AI Claude Code for Hardware • PRD to 3D Enclosure

Ohm — Autonomous AI Agent for Hardware & IoT Engineering

Lead Hardware AI Systems Architect · Autonomous Hardware Agent ('Claude Code for Hardware')

Source Code Watch Demo →

Architecture & Implementation

Autonomous hardware engineering platform that functions as 'Claude Code for hardware engineers'. Interactively deliberates with the user to construct an MVP PRD and SRS document, autonomously fetches real-world IoT components and microcontrollers from Amazon and Flipkart with pin-compatibility and live pricing checks, auto-wires circuit schematics in KiCAD with SPICE simulation verification, and controls Blender/CAD to generate production-ready 3D bodies and snap-fit enclosures.

✓ What I Learned From It

Engineered multi-modal computer control loops for CAD/EDA software, automated headless scraping for component pinout matching, and bridged high-level PRD requirements into mechanical 3D geometry.

✓ What I Now Know

How to automate the entire hardware development cycle from natural language ideation to KiCAD PCB layout, BOM part procurement, and Blender 3D enclosure modeling.

✓ Core Experience Gained

Autonomous hardware agents, computer control, KiCAD schematic routing, Blender Python API, automated BOM sourcing, and IoT system architecture.

TECH STACK: Python Computer Use KiCAD Automation Blender API IoT Systems Autonomous Agents
2026 CYBERSECURITY IIT Hyderabad Valor Award • ₹60,000+ Prize • AST Graph

Sentinel — Banking Fraud Detection & APK Reverse Engineering Pipeline

Lead Security Architect & Full-Stack Developer · Bank of India Hackathon Winner

Source Code Watch Demo →

Architecture & Implementation

Automated Android APK reverse engineering and banking fraud vulnerability scanner that won the IIT Hyderabad hackathon (₹60,000+ prize, Special Valor Award). Decompiles production APK bytecode into readable Java through an automated 8-tool pipeline (integrating MobSF, JADX, and decompilers). Solved the LLM context-window limit by constructing an AST (Abstract Syntax Tree) Knowledge Graph of the entire codebase for AI reasoning and RBI-compliant audit reporting. Incorporates Bank Anti-Malware analysis tooling.

✓ What I Learned From It

Automated manual 8-tool reverse engineering pipelines and engineered AST Knowledge Graphs to overcome LLM context-window bottlenecks when analyzing massive codebases.

✓ What I Now Know

How to decompile obfuscated Android APKs into readable Java, map repository AST structures for AI reasoning, and generate automated RBI-compliant security audit reports.

✓ Core Experience Gained

Enterprise mobile security, APK reverse engineering (MobSF, JADX), AST parsing, knowledge graphs, and automated financial compliance.

TECH STACK: Python FastAPI MobSF JADX AST Knowledge Graphs PostgreSQL
2026 CLIENT 100% Self-Hosted • Zero Cloud Fees • Instant Lead Triage

n8n Automation — Multi-Channel Lead Triage & WhatsApp CRM

Automation Architect & Consultant · SaaS & Retail Clients (India & Dubai)

Source Code Live Platform →

Architecture & Implementation

Fully self-hosted event-driven sales automation pipeline running on private infrastructure with zero cloud/server fees. Combines Meta WhatsApp Cloud API and email bots to capture inbound leads, score intent via Groq, update CRM records, and trigger instant messaging. Integrated with content systems like Vyx for automated lead conversion.

✓ What I Learned From It

Engineered webhooks, multi-channel state machine routing in n8n, self-hosted deployment models, and WhatsApp Cloud API session management.

✓ What I Now Know

How to run enterprise automation pipelines 100% free on self-hosted infra, eliminating recurring SaaS subscription fees for clients.

✓ Core Experience Gained

n8n workflow architecture, WhatsApp Cloud API, webhook triggers, CRM data synchronization, and automated lead scoring.

TECH STACK: n8n (Self-Hosted) WhatsApp API Webhooks Node.js Groq PostgreSQL
2026 AI Sub-Second Voice Latency • 1-2 Min Bot Creation

Bol — No-Code Generative AI Voice Agent Builder

Lead AI Architect & Founder · Generative Voice AI Platform

Watch Demo →

Architecture & Implementation

No-code conversational AI studio that lets founders build, test, and deploy interactive voice agents in minutes through natural speech. Features sub-second turn-taking, multi-language Indian dialect synthesis on Gnani.ai models, custom tool-calling, and zero manual code.

✓ What I Learned From It

Implemented low-latency streaming TTS/STT pipelines and stateful voice turn-taking with interrupted speech recovery.

✓ What I Now Know

How to design human-sounding conversational voice loops with sub-second response times using Gnani speech models and WebSockets.

✓ Core Experience Gained

Conversational voice AI, WebSockets audio streaming, FastAPI backends, and low-latency speech interfaces.

TECH STACK: Next.js Voice AI Gnani.ai WebSockets FastAPI Speech Synthesis
2026 AI IDBI Hackathon Finalist • Autonomous Modeling

IDBI Wealth Engine — AI Wealth Advisory & Portfolio Engine

Full-Stack AI Engineer · IDBI Hackathon / Wealth Advisory

Source Code Live Platform →

Architecture & Implementation

AI-driven investment management and automated wealth advisory engine designed for the IDBI Hackathon, orchestrating real-time portfolio balancing, government scheme eligibility checks, and multilingual voice assistance for semi-urban and rural customers.

✓ What I Learned From It

Learned quantitative financial modeling, risk-weighted portfolio optimization, and high-throughput LLM advisory pipelines.

✓ What I Now Know

How to design accessible fintech advisory workflows, integrate real-time market data, and produce algorithmic asset allocations for diverse populations.

✓ Core Experience Gained

FinTech advisory engines, financial risk modeling, FastAPI microservices, and AI-driven wealth management.

TECH STACK: Next.js Python FastAPI LLMs Voice AI FinTech
2026 CYBERSECURITY SIH 2026 Cyber Intelligence • Entity Resolution

Jane Intelligence — SIH 2026 Dark Web Investigation Pipeline

Lead Backend & Intelligence Architect · Smart India Hackathon (SIH 2026)

Source Code Live Platform →

Architecture & Implementation

Dark web de-anonymization investigation platform and entity graph resolution pipeline for cyber intelligence, featuring multi-hop entity resolution and forensic timeline indexing.

✓ What I Learned From It

Architected multi-node graph databases for entity resolution, asynchronous onion network scrapers, and timeline correlation models.

✓ What I Now Know

How to de-anonymize illicit threat actors across disparate intelligence silos and present actionable evidence graphs.

✓ Core Experience Gained

Cyber intelligence platforms, graph network indexing, forensic law enforcement tools, and FastAPI backend engineering.

TECH STACK: FastAPI Python Graph Networks Next.js Vercel
2026 AI Autonomous 3D Rocket Generation • Any-LLM Computer Use

ARIA AI Platform — Embodied Voice & Multimodal Computer Control

System Architect & Lead Developer · Autonomous AI Research

Source Code Watch Demo →

Architecture & Implementation

Embodied AI platform featuring low-latency real-time voice streaming, multimodal perception, and an 8-agent desktop control framework with Redis shared memory. Autonomously engineered a 3D model of a rocket directly on desktop CAD software using a compact local DeepSeek-style model, outperforming heavy frontier models on latency and efficiency.

✓ What I Learned From It

Engineered low-latency WebSockets for bidirectional voice streaming and multi-agent coordination with Redis-backed shared state.

✓ What I Now Know

How to build autonomous agentic loops that control operating systems and desktop software with high precision on lightweight local models.

✓ Core Experience Gained

Autonomous agent architectures, Redis shared memory, computer vision OS automation, and real-time voice AI.

TECH STACK: Python Docker Redis Multi-Agent Systems VNC Automation DeepSeek
2024-Present CLIENT ChatGPT & Claude Recommended • UAE Client Reach

RafRaf International — Global Import & Export Platform

Freelance Full-Stack Developer & SEO Architect · RafRaf International (UAE Client)

Source Code Live Platform →

Architecture & Implementation

End-to-end production web platform built for a UAE-based international trade client. Features Next.js 16 architecture, Figma design, structured catalog indexing, and comprehensive SEO/AEO/GEO optimization that causes ChatGPT and Claude to actively recommend the client for global trade.

✓ What I Learned From It

Handled full client lifecycle from Figma design to production Next.js deployment, international SEO, and brand storytelling.

✓ What I Now Know

How to ship production B2B platforms that rank on Google, get cited by AI answer engines, and convert international buyers.

✓ Core Experience Gained

End-to-end B2B client delivery, international SEO/AEO/GEO, Next.js architecture, and commercial lead generation.

TECH STACK: Next.js React TypeScript Tailwind CSS Figma SEO/AEO/GEO
2026 AI Self-Hostable • Custom Domains & Gmail • Autonomous Triage

Agentic Inbox — Autonomous Email Intelligence & Triage

Founder & AI Engineer · Autonomous Productivity Tools

Source Code Live Platform →

Architecture & Implementation

Self-hostable AI email client and operations engine supporting Gmail, IMAP/SMTP, and any custom domain address (@yourdomain.com). Orchestrates autonomous email categorization, automated draft replies, smart meeting scheduling, and semantic context search.

✓ What I Learned From It

Mastered semantic clustering of unstructured text, OAuth token handling at scale, and prompt caching to cut inference costs.

✓ What I Now Know

How to build reliable autonomous email agents that classify, prioritize, and draft executive responses without human intervention across custom domains.

✓ Core Experience Gained

Productivity AI agents, LangChain/OpenAI integrations, email parsing infrastructure, and API authentication.

TECH STACK: TypeScript Next.js LangChain OpenAI / Claude Custom SMTP/IMAP
2025-2026 AI 0.03ms Cached Latency • Zero API Calls • Refuses to Lie

HHGoa In-House RAG — Sub-Millisecond Multilingual LLM Pipeline

AI Systems Architect · HHGoa Hackathon Submission

Source Code Live Platform →

Architecture & Implementation

Custom-built in-house RAG pipeline with zero external API calls. Features locally deployed LLM, prompt caching yielding sub-millisecond retrieval (0.03 ms cache hits to 1,473 ms cold queries), custom safety guardrails, native Hindi multilingual responses, and strict anti-hallucination thresholds that refuse to lie.

✓ What I Learned From It

Engineered ultra-low-latency in-memory prompt caching, local LLM quantization, and strict epistemic verification guardrails.

✓ What I Now Know

How to achieve 0.03 ms response speeds on local hardware with zero external cloud dependencies while guaranteeing truthfulness.

✓ Core Experience Gained

Local LLM deployment, RAG pipeline architecture, prompt caching, multilingual NLP (Hindi), and guardrail engineering.

TECH STACK: Local LLM RAG Pipeline Prompt Caching Hindi NLP Guardrails Python
2025-2026 WEB IIT Bhubaneswar Winner • Gamified Habit RPG

Quest — Gamified Life RPG & Habit Mastery Engine

Founder & Full-Stack Architect · IIT Bhubaneswar Hackathon Winner

Source Code Live Platform →

Architecture & Implementation

RPG-style, gamified life and habit tracking web application that won top honors at the IIT Bhubaneswar Web Hackathon. Turns daily routines and tasks into character progression, skill trees, and real-time XP leveling. Built with Next.js 14, Supabase, PostgreSQL, and Framer Motion.

✓ What I Learned From It

Engineered real-time gamification state machines, habit streak algorithms, and fluid RPG interface dynamics with Framer Motion.

✓ What I Now Know

How to apply behavioural psychology and RPG game loops to productivity tools to maximize daily user retention.

✓ Core Experience Gained

Gamification architecture, Supabase PostgreSQL, Next.js 14, Framer Motion animations, and full-stack product design.

TECH STACK: Next.js 14 Supabase PostgreSQL Framer Motion Tailwind CSS
2025-2026 WEB FIFA Hackathon • Real-Time 3D Wayfinding

FIFA 3D Stadium Map — Interactive Tournament Navigation

Creative 3D Developer · FIFA Hackathon Submission

Source Code Live Platform →

Architecture & Implementation

Interactive 3D stadium wayfinding and spatial navigation platform built for the FIFA Tournament Hackathon. Solves fan navigation challenges by providing multi-tiered 3D digital twins of tournament stadiums, guiding spectators to exact seating blocks, concourses, and transit gates in real-time WebGL 3D.

✓ What I Learned From It

Created 3D architectural digital twins optimized for web browsers and mobile GPU rendering with minimal memory footprint.

✓ What I Now Know

How to build high-performance spatial wayfinding applications with interactive camera navigation and section filtering.

✓ Core Experience Gained

Three.js, WebGL spatial mapping, interactive camera staging, React Three Fiber, and 3D asset optimization.

TECH STACK: Three.js WebGL Next.js React Three Fiber 3D Mapping
2026 CYBERSECURITY Legacy CCTV AI Retrofit • ALPR & Facial Recognition

Smart CCTV & Border Surveillance — Perimeter Threat AI

AI Computer Vision Engineer · SIH 2026 PS-187 (Border Defense)

Source Code Live Platform →

Architecture & Implementation

AI computer vision retrofit system engineered for high-security Indian border forts and military outposts. Connects directly to existing legacy CCTV feeds to add real-time facial recognition, automated license plate recognition (ALPR/ANPR) for every entering/exiting vehicle, personnel entry/exit tracking, and instant alerts for suspicious perimeter movements.

✓ What I Learned From It

Trained YOLO object detection pipelines on edge hardware with RTSP stream ingestion, automated plate extraction, and low-latency inference.

✓ What I Now Know

How to retrofit non-smart legacy camera hardware with modern AI capabilities and real-time defense alerting.

✓ Core Experience Gained

Edge computer vision, YOLOv8 tracking, video pipeline optimization, ALPR/ANPR, and defense-grade perimeter alerts.

TECH STACK: Python Computer Vision YOLOv8 ALPR/ANPR FastAPI CCTV AI
2025 CLIENT AWS Event Founder Pitch • High-Converting Redesign

Ledgerpe — High-Converting FinTech Web Redesign

Frontend UI/UX Designer & Developer · FinTech Founder Pitch (AWS Event)

Source Code Watch Demo →

Architecture & Implementation

High-impact commercial web redesign built after meeting the founder at an AWS industry event. Transformed their outdated legacy web interface into a sleek, high-converting modern FinTech landing experience with fluid typography, responsive components, and modern conversion architecture (currently in founder review and pitch phase).

✓ What I Learned From It

Analyzed commercial user dropoff on legacy FinTech sites and engineered high-converting modern landing page hierarchies.

✓ What I Now Know

How to translate executive founder requirements into conversion-engineered web designs that establish immediate trust.

✓ Core Experience Gained

Fintech landing page redesign, conversion rate optimization (CRO), React 18, and modern typography.

TECH STACK: React 18 Next.js TypeScript Tailwind CSS UI/UX Redesign
2025-2026 AI Multi-Agent Consensus Debate • Concurrent Tri-Model UI

Multi-LLM Assistant — Collaborative Consensus & Streaming Interface

Full-Stack Developer · Developer Tools

Source Code Live Platform →

Architecture & Implementation

Multi-AI collaborative consensus interface where multiple LLMs (Claude, GPT, Gemini) respond simultaneously and debate among themselves in real time, cross-examining arguments to synthesize the most accurate, high-depth answer for complex queries.

✓ What I Learned From It

Unified disparate LLM streaming APIs into a standard multi-agent debate pipeline with consensus synthesis.

✓ What I Now Know

How to orchestrate multiple competing AI models simultaneously with cross-examination and latency telemetry.

✓ Core Experience Gained

Multi-model AI streaming, multi-agent debate architectures, token optimization, and developer tooling.

TECH STACK: Next.js React TypeScript Multi-Agent Debate Claude API GPT-4 Gemini
2025-2026 AI Semantic Regulatory Search • Automated Workflows

Politiks — Civic Intelligence & Legal Tech RAG Platform

AI Engineer · Civic Intelligence

Source Code Live Platform →

Architecture & Implementation

Civic intelligence and legal technology platform utilizing retrieval-augmented generation (RAG) to query complex legislation, court records, and civic regulatory documents with automated citation verification.

✓ What I Learned From It

Tuned chunking strategies and hybrid dense-sparse vector search on legal jurisprudence and municipal statutes.

✓ What I Now Know

How to build high-precision RAG pipelines where hallucinations are minimized through strict citation verification.

✓ Core Experience Gained

RAG architectures, vector database search (Chroma/Pinecone), legal document processing, and FastAPI backends.

TECH STACK: Python FastAPI Vector DB RAG Next.js
2025 WEB Top 40 / 2,000+ Students • 98.4% R² Score

ML Reactor Yield Engine — Chemical Kinetics Ensemble Regression

ML Researcher & Data Scientist · IIT ML Hackathon (Top 40 / 2,000+)

Source Code Live Platform →

Architecture & Implementation

Machine learning regression pipeline developed for an IIT hackathon, ranking in the Top 40 out of 2,000+ participating engineering students. Predicts chemical reactor yield outcomes under varying physical and catalytic conditions with 98.4% R² test accuracy using ensemble XGBoost and Scikit-learn algorithms.

✓ What I Learned From It

Extracted signal from noisy chemical process sensor data using ensemble regression models and hyperparameter cross-validation under competition constraints.

✓ What I Now Know

How to clean real-world multivariate tabular data and build highly accurate predictive regression pipelines outperforming 1,960+ peers.

✓ Core Experience Gained

Tabular data engineering, predictive modeling with XGBoost, cross-validation, and production model evaluation.

TECH STACK: Python Scikit-Learn XGBoost Pandas Jupyter
2025-2026 WEB 6,000+ Peak Users • 4,500–5,000 MAU

Awesome LM — NotebookLM Presentation & PPT Extension

Creator & Lead Engineer · Productivity & AI Tools

Source Code Live Platform →

Architecture & Implementation

Presentation design companion inspired by Google NotebookLM. While NotebookLM generates content from user notes, its default slide aesthetics are limited. Awesome LM delivers polished, executive-ready presentation templates and themes to NotebookLM workflows via structured prompts and a dedicated Chrome extension. Reached 6,000+ peak concurrent users at launch, maintaining 4,500–5,000 monthly active users.

✓ What I Learned From It

Engineered Chrome extension architecture for DOM manipulation, prompt template injection, and multi-format slide export.

✓ What I Now Know

How to build consumer developer tools that solve specific friction in Google products and maintain thousands of active users.

✓ Core Experience Gained

Chrome extension development, prompt engineering, TypeScript, NotebookLM ecosystem, and consumer product growth.

TECH STACK: Chrome Extension Prompt Engineering TypeScript Tailwind CSS NotebookLM
2025 WEB 60 FPS Animation • Physics Spring Dynamics

Awwwards Nav Rebuild — Fluid Motion Interaction Lab

Frontend Motion Engineer · Creative Frontend Lab

Source Code Live Platform →

Architecture & Implementation

Pixel-perfect rebuild of award-winning website navigation architectures featuring physics-based spring curves, magnetic button hovers, dynamic SVG paths, and 60 FPS transitions.

✓ What I Learned From It

Pushed Framer Motion and GSAP to their limits, eliminating layout reflows and keeping paint times under 16ms.

✓ What I Now Know

How to craft bespoke, award-caliber UI interactions that make high-end products feel tactile and expensive.

✓ Core Experience Gained

Creative frontend engineering, GSAP/Framer Motion physics curves, SVG path morphing, and 60 FPS UI performance.

TECH STACK: React Framer Motion GSAP Tailwind CSS
2025-2026 WEB Interactive 3D Experience • WebGL Powered

Nabil Portfolio 2.0 — 3D Web Experience & Gallery

Creative Developer & 3D Artist · Creative Engineering

Watch Demo →

Architecture & Implementation

Interactive 3D web experience built with Next.js, Three.js, and GLSL shaders, featuring dynamic camera staging, interactive 3D assets, and photography gallery.

✓ What I Learned From It

Wrote custom vertex and fragment shaders in GLSL, optimized WebGL render loops, and managed mobile GPU battery drain.

✓ What I Now Know

How to blend 3D graphics seamlessly with standard React web layouts for immersive commercial storytelling.

✓ Core Experience Gained

Three.js & React Three Fiber, GLSL shader authoring, 3D asset optimization, and creative web direction.

TECH STACK: Next.js Three.js React Three Fiber GLSL Shaders
2024-2025 WEB 2,000+ Organic Monthly Active Users

Gitskinz — Developer Profile Architecture & Generator

Founder & Solo Creator · Developer Tools (Gitskinz)

Source Code Live Platform →

Architecture & Implementation

Developer platform featuring 70+ brutalist and minimalist GitHub README profile architectures. Organically scaled to 2,000+ monthly active users with zero paid marketing.

✓ What I Learned From It

Discovered product-market fit through organic developer communities, markdown template synthesis, and viral sharing mechanics.

✓ What I Now Know

How to build and launch a developer product that acquires thousands of real users with zero ad spend.

✓ Core Experience Gained

Product bootstrapping, developer community growth, zero-budget user acquisition, and lightweight web architecture.

TECH STACK: Next.js React Vite Tailwind CSS
2018-2019 AI Built in 2018 • Early OpenAI Pioneer • 8+ Yrs AI

Nbail 2018 Chatbot & Agent — Pioneer AI Project (8+ Yrs Experience)

Early AI Pioneer & Developer · Pioneer Hackathon Winner

Live Platform →

Architecture & Implementation

Nabil's first artificial intelligence project, engineered in 2018 as one of the earliest adopters of OpenAI's APIs before modern ecosystems existed. Created a conversational assistant with primitive computer-use agentic capabilities and system task automation that earned early hackathon victories and established Nabil's 8+ years of hands-on experience in generative AI and agentic systems.

✓ What I Learned From It

Mastered prompt engineering and raw HTTP API orchestration when no wrappers, frameworks, or cloud agent ecosystems existed.

✓ What I Now Know

How generative AI systems have evolved from early completions to modern multi-agent autonomous computer control.

✓ Core Experience Gained

Pioneer AI exploration, early OpenAI completions, command-line automation, and 8+ years long-term domain insight.

TECH STACK: Early OpenAI APIs Python Computer-Use Automation Conversational AI