About Skills Projects AI Work Journey Contact Hire Me
AI Engineer  ·  Builder  ·  Founder

Saharsh Kashyap

Building the Future of Artificial Intelligence

I am a 2nd-year B.Tech CSE student, AI engineer, and founder of TrivoX Technologies. I build production-grade LLM systems, RAG pipelines, and AI products from scratch — while pursuing a mission to make intelligence accessible for everyone.

🏛️ JEC Jabalpur · Batch 2028 🏐 National Volleyball Player 🚀 Founder · TrivoX Technologies 🧠 AI Engineer Since 2023
🧠 LLM Engineer
⚡ Founder · TrivoX
🏐 National Athlete
About Me

Rooted in Purpose,
Driven by Vision

I'm Saharsh Kashyap from Jabalpur, Madhya Pradesh — a second-year B.Tech Computer Science student at Jabalpur Engineering College (JEC), Batch 2028.

My journey into AI began in 2023, when I started self-learning LLMs, fine-tuning, and building real AI products from scratch. I'm implementation-first by nature — I learn by building, not by watching from the sidelines.

I founded TrivoX Technologies, a registered software company, with a long-term mission to make AI and technology accessible regardless of economic background. I believe intelligence should not be a privilege.

Outside of code, I'm a national-level volleyball player, which taught me discipline, teamwork, and the ability to perform under pressure — skills I bring to every product I build.

My target: become a world-class AI engineer by 2028, contribute to frontier AI research, and build products that matter for millions.

2+
Years Building AI
3+
Live AI Products
10+
Clients Served
3K+
LinkedIn Network
Life Highlights

The Layers
Behind the Code

🏐

National Volleyball Player

Competed at the national level — discipline, coordination, and relentless practice define both my game and my engineering approach.

🏛️

JEC Jabalpur · CSE 2028

Pursuing B.Tech in Computer Science while simultaneously building two companies and executing a 24-month AI engineering roadmap.

🕉️

Vedic Astrology & Spirituality

Deep interest in Vedic thought and devotional practice — keeping me grounded as I navigate the fast-moving world of technology.

🌍

Humanitarian Vision

Core mission: make world-class education and technology accessible to every student regardless of their economic background.

Technical Arsenal

Skills &
Expertise

🧠 AI & Machine Learning

LLMsRAG PipelinesFine-Tuning LangChainHuggingFaceEmbeddings Vector DatabasesPrompt Engineering

⚙️ Backend & Infrastructure

FastAPIPythonNode.js SQLite / PostgreSQLREST APIsDocker

🎨 Frontend & Products

ReactNext.jsHTML/CSS/JS Three.jsSaaS ArchitectureNo-Code Platforms

🔬 Deep Learning & Research

Neural NetworksTransformersAttention Mechanisms Deep Learning TheoryNPU/GPU Inference

Proficiency
Levels

LLM Development & RAG90%
FastAPI / Python Backend88%
AI Product Development85%
React / Frontend78%
Deep Learning Theory72%
ML Research & Experimentation68%
System Architecture80%
What I've Built

Projects & Products

Every line of code serves a purpose. These are the products I've built, shipped, and scaled — from idea to production.

S
Sakhi AI Screenshot Launching Soon

Sakhi AI

Hindi-first AI assistant for RGPV university students. Full RAG pipeline seeded with real syllabus, past year questions, and study material. Answers queries in native language 24/7 with sparkle-based credit monetization.

FastAPIRAGSQLiteHindi NLPLLMB2C
C
ChatBuilder Screenshot Building

TrivoX ChatBuilder

No-code AI chatbot builder for Indian SMBs. Full B2B SaaS with dark-themed dashboard UI, multi-LLM backend, custom knowledge base training, and pricing tiers from ₹999 to enterprise. My primary revenue play for 2025.

No-Code SaaSMulti-LLMReactFastAPIB2B
E
School ERP Screenshot Live

School ERP System

Comprehensive school management platform live in production. Handles admissions, student records, attendance tracking, fee management, staff HR, and analytics — built from scratch for the Indian education system.

ERPMulti-roleAnalyticsFull-Stack
V
VRYKON Screenshot Research

VRYKON Runtime

NPU-first AI inference runtime — an Ollama alternative with intelligent NPU/dGPU/iGPU routing. Designed for Ryzen AI hardware (50 TOPS NPU). Open-source vision to democratize local AI inference. Long-term mission project.

NPUInference RuntimeOpen SourceC++/PythonRyzen AI
M
Marketing Site Screenshot Delivered

Marketing Agency Website

Full professional website built for a marketing company client. Modern responsive design, SEO optimized, custom CMS integration — delivered as part of TrivoX's client services portfolio.

ReactSEOResponsiveClient Work
T
TrivoX Screenshot Founded

TrivoX Technologies

My registered software company — providing AI chatbots, web apps, and ERP systems to Indian and international clients. Building toward a vision of accessible, intelligent software for every business.

FounderRegistered Co.AI ServicesSaaS Products
AI & Machine Learning

What I Know & What I'm Building

I've been working with LLMs and AI systems since 2023 — not just using them, but engineering them. Here's the depth of what I know and where I'm heading.

01
🧠

Large Language Models

Deep understanding of transformer architecture, attention mechanisms, tokenization, and inference. I've worked with GPT-4, Gemini, Mistral, LLaMA, and custom fine-tuned models. I understand both the magic and the math.

GPT-4GeminiMistralLLaMATransformers
02
📚

RAG Systems

Built multiple production RAG pipelines from scratch. Document ingestion, chunking strategies, embedding models, vector store selection (FAISS, Chroma), retrieval ranking, and response synthesis — I've engineered every layer.

LangChainFAISSChromaEmbeddingsChunking
03
🎯

Fine-Tuning & Training

Hands-on experience with model fine-tuning using LoRA/QLoRA, PEFT techniques, and instruction tuning. Currently deep-diving into Andrej Karpathy's Zero to Hero and building micrograd-style implementations to understand training from scratch.

LoRAQLoRAPEFTInstruction Tuning
04

AI Inference & Hardware

Exploring NPU-first inference on Ryzen AI 7 350 (50 TOPS). Researching intelligent hardware routing between NPU, dGPU (RTX 5050), and iGPU (Radeon 880M). Building VRYKON as the runtime that makes local AI truly efficient.

NPURTX 5050Ryzen AIOllama Alt
05
🔬

Deep Learning Theory

Actively studying 3Blue1Brown neural networks, MIT Deep Learning, and Stanford CS231n. Understanding backpropagation, gradient descent variants, loss landscapes, and architectural choices — not just using models, but understanding why they work.

BackpropCNNRNNAttentionOptimization
06
🚀

AI Product Engineering

I build AI products end-to-end — from the LLM backbone to the user interface. FastAPI backends, vector stores, credit systems, multi-tenant SaaS, and production deployment. AI without a product is just research — I ship both.

FastAPIProduction LLMSaaSMulti-tenant
07
🗣️

Multilingual & Hindi NLP

Built Sakhi AI with a Hindi-first NLP pipeline tailored for RGPV university students — handling Hinglish, code-mixing, and domain-specific education queries. Explored Sarvam AI and IndicBERT for Indian language understanding at production scale.

Hindi NLPHinglishSarvam AIIndicBERTCode-mixing
08
🔧

Local AI & Ollama Stack

Running a full local AI assistant stack on HP Omen (Ryzen AI 7 350, RTX 5050) using Qwen2.5-Coder-14B via Ollama, Open WebUI, and SearXNG with a custom Hinglish system prompt. Studying llama.cpp internals as the foundation for VRYKON's runtime engine.

OllamaQwen2.5Open WebUIllama.cppLocal LLM
Active Learning Roadmap

24-Month AI Engineering Roadmap (2026–2028)

Track 01
ML/DL Theory
3Blue1Brown · Karpathy Zero to Hero · HuggingFace Course · MIT Deep Learning
Track 02
Practical Projects
Sakhi AI · ChatBuilder · VRYKON · Client Products · Open Source Contributions
Track 03
Research Internships
Targeting: Sarvam AI · Krutrim · IIIT-H NLP Lab · IIT Research Collaborations
Track 04
VRYKON — Open Source
NPU-first inference runtime · AI-first Linux distro · Community traction target: 2026
My Story

The Journey
So Far

From Jabalpur to building AI systems used by real people — this is how it unfolded.

2020–2022

The Beginning — Code & Volleyball

Started programming while competing at national level volleyball. Discovered that discipline from sports translated directly into engineering. Built first websites and learned fundamentals of software development.

2023

AI Awakening — LLMs & First Products

Discovered LLMs and immediately started building. Self-taught prompt engineering, fine-tuning, and RAG systems. Built first AI-integrated products and realized this was my calling — making machines think.

2024 · JEC

Joined B.Tech CSE — Building in Parallel

Enrolled at Jabalpur Engineering College (JEC), Batch 2028. Immediately started building while studying — Sakhi AI for RGPV students, School ERP system, and marketing websites. Registered TrivoX Technologies officially. Built 3000+ LinkedIn network.

2025

Scaling — ChatBuilder, VRYKON & Global Clients

Executing 24-month roadmap: shipping ChatBuilder as primary revenue engine, developing VRYKON open-source, targeting Sarvam AI internship, and expanding into international freelance market via Fiverr and direct outreach.

2026–2028 · Vision

World-Class AI Engineer — The Mission

Target: VRYKON achieves community traction, research internship at a top Indian AI lab, strong international freelance income, and becoming one of India's most impactful young AI engineers by graduation. The mission is just beginning.

Get In Touch

Let's Build
the Future
Together

Whether you have a project, a collaboration idea, or just want to talk AI — I respond to every serious message within 24 hours.

📍
Location
Jabalpur, Madhya Pradesh, India
🎓
Education
Jabalpur Engineering College · CSE · Batch 2028
💳
Available For
Freelance · Internships · Collaborations