const darshan = {
name: "Darshan Rajput",
role: "AI Engineer & Open Source Maintainer",
education: "B.Tech Computer Engineering @ Marwadi University",
location: "Gujarat, India",
ai_architecture: [
"Real-Time Voice AI Agents (Sub-2s Latency)",
"Multi-Agent RAG & Graph Memory Systems (Cognee / Vector DBs)",
"Multi-LLM Routing & Evaluation Pipelines",
"Autonomous Workflow Automation (n8n / Tool Calling)"
],
maintaining: "The Dev Pocket — 100+ contributors, 150+ PRs reviewed",
stack: [
"Next.js", "TypeScript", "Python",
"OpenAI / Claude / Gemini", "Vapi AI", "AssemblyAI", "n8n"
],
open_to: [
"AI/ML Engineering Internships",
"Forward Deployed Engineer (AI)",
"Full Stack AI Roles",
"High-Growth AI Startups"
]
};- 🤖 Architecting production LLM systems, real-time voice AI agents & autonomous pipelines
- 🧠 Building AI memory connectors & knowledge graphs (Top PR Winner @ Cognee AI Hackathon)
- 🛠️ Project Admin & Lead Maintainer at The Dev Pocket — 100+ contributors, 150+ PRs reviewed
- 🌍 Active Open Source Leader in GSSoC 2026, ECWoC 2026, and Hacktoberfest 2024 & 2025
- 🎯 Competed in Code Carnival S02 — National 36-hour Hackathon at Atmiya University
Moving beyond simple prompt wrappers—designing low-latency, fault-tolerant AI systems that integrate cleanly with production data and enterprise workflows.
flowchart LR
A[🎙️ Voice / Text Input] -->|WebSockets / Vapi| B(⚡ Real-Time STT & Intent Router)
B -->|Multi-LLM Orchestration| C{🧠 AI Agent Core}
C <-->|DLT / Vector & Graph Memory| D[(📚 Cognee / NeonDB)]
C -->|Deterministic Tool Calling| E[⚙️ n8n / Calendar / APIs]
C -->|Streaming TTS| F[🔊 Sub-2s Voice Response]
| AI Engineering Pillar | Production Implementation & Focus |
|---|---|
| 🎙️ Real-Time Voice AI | End-to-end conversational pipelines using Vapi AI, AssemblyAI, and streaming LLMs with sub-2-second response latency and automated turn-taking. |
| 🧬 AI Memory & RAG Pipelines | Incremental data ingestion using DLT (Data Load Tool), semantic chunking, forget-on-delete synchronization, and knowledge-graph memory (Cognee). |
| 🔀 Multi-LLM Orchestration | Unified interfaces comparing OpenAI, Anthropic Claude, and Google Gemini side-by-side with structured CRISPE prompt engineering. |
| ⚡ Agentic Workflows | Autonomous classification, intent extraction, and multi-step tool execution via n8n, OpenAI Function Calling, and webhook pipelines. |
Building in public, shipping production fixes to high-star repositories, and collaborating with core maintainers across the AI and data ecosystem.
WeMakeDevs × OpenMetadata (Back to the Metadata) & Cognee (The Hangover Part AI)
| 🔥 Contributions | 📦 Repositories | 🐛 Issues Resolved | 🔀 Merged PRs | 🏅 Hackathon Prize |
|---|---|---|---|---|
| Production Engineering | 3 | 4 | 4 | $100 USD (Top PR Winner) |
Where's My Context? · Cognee × WeMakeDevs
🏆 Top PR Winner · Cognee AI Memory Hackathon · WeMakeDevs
🚀 Creator & Lead Maintainer of The Dev Pocket — an open-source developer platform containing curated resources, interactive tools, learning material, and developer roadmaps.
| 🔁 PRs Reviewed | 👥 Contributors | 🌍 Open Source Programs | ⭐ Active Since |
|---|---|---|---|
| 150+ | 100+ | GSSoC · Hacktoberfest · ECWoC | Oct 2025 |
| Role | Program | Duration | Engineering & Community Impact |
|---|---|---|---|
| Project Admin | ECWoC 2026 — Elite Coders | Jan 2026 – Present | Led 100+ contributors · 30% increase in active PR velocity |
| Lead Maintainer | Hacktoberfest 2025 — DevPocket | Oct 2025 – Jan 2026 | Reviewed & merged 150+ PRs · Automated CI/CD & PR triage |
| Open Source Contributor | GSSoC 2024 + Hacktoberfest 2024 | Oct – Nov 2024 | Shipped features & fixes across 8+ production repositories |
|
Real-time conversational healthcare AI platform powered by low-latency speech recognition and clinical LLM prompting. 📈 40% accessibility improvement · 1,000+ concurrent users
|
Autonomous voice receptionist handling inbound customer calls, intent detection, live calendar lookup, and booking workflows. ⚡ Zero manual scheduling · Sub-2s voice latency
|
|
Unified multi-model AI playground enabling real-time side-by-side response benchmarking across frontier LLMs. 🚀 35% higher user session engagement
|
LLM-powered email pipeline that classifies inbound threads, extracts urgency/intent, and drafts context-aware replies. ⏱️ 60% reduction in manual inbox processing
|
|
Chrome extension that automatically restructures raw user queries into engineered CRISPE-framework system prompts. 🔒 Zero-config · Client-side execution
|
Cross-platform mobile application featuring interactive emission analytics and personalized sustainability recommendations. 🏅 Code Carnival S02 National Hackathon Entry
|
🎖️ View Open Source Badges (Layer5, GSSoC, Hacktoberfest & Holopin)
📈 View Detailed GitHub Streak, Activity Graph & LeetCode Heatmap
| Key Milestone | |
|---|---|
| 🤖 | Architected real-time Voice AI systems serving 1,000+ concurrent users with sub-2s latency |
| 🏆 | Top PR Winner ($100 Prize) — Built a production Confluence → Cognee AI Memory DLT connector |
| 🎖️ | Back to the Metadata Hackathon 2026 — Merged production K8s & UTF-8 PRs into OpenMetadata & Lingo.dev |
| 🔁 | Reviewed, triaged & merged 150+ Pull Requests as Lead Maintainer of The Dev Pocket |
| 👥 | Mentored 100+ global open-source contributors across ECWoC, GSSoC & Hacktoberfest |
| 🚀 | Shipped 6+ end-to-end AI products spanning Voice Agents, Multi-LLM Arenas, and Autonomous Workflows |
AI/ML Engineering Internships • Forward Deployed Engineer (AI) • Full Stack AI Roles • AI Startups • Open Source Collaborations
⭐ Built & Maintained by Darshan Rajput