Hi, I'm

Darsh Bothra

Applied AI Engineer • LLMs • Computer Vision • Backend Systems

I build production AI systems — from LLM-powered backends to computer vision on edge devices. Currently seeking roles where I can ship reliable AI products end to end.

View ProjectsDownload Resume

About

I'm a final-year Electronics & Communication Engineering student at NIT Surat, focused on AI engineering and backend systems. I work with LLMs, computer vision models, FastAPI, and retrieval pipelines — shipping systems that run in production, not just notebooks. I care about inference speed, clean APIs, and products people can actually use.

Experience

  1. Jan 2026 – Jul 2026

    Indore, Madhya Pradesh

    AI Engineer Intern

    Quasi Intelligence

    • Fine-tuned RF-DETR, RT-DETR, and lightweight CNNs on 100K+ images for industrial object detection and classification.
    • Cut inference latency by exporting PyTorch models to ONNX and TensorRT (FP16/INT8) for Jetson Orin Nano and Raspberry Pi 5.
    • Built NVIDIA DeepStream pipelines for 6+ concurrent RTSP streams, plus a GraphRAG document system with LangGraph, Qdrant, and Ollama.

  2. May 2025 – Jul 2025

    Ontario, Canada (Remote)

    Software Development Intern — Automation & AI

    Shield Identity

    • Built n8n automation workflows to scrape and categorize AI & cybersecurity news via RSS feeds, storing structured daily, weekly, and yearly summaries in a database.
    • Explored LangChain for LLM-based summarization and retrieval workflows.
    • Automated WordPress content publishing via n8n for seamless blog updates.

Projects

Selected work in AI products, ranking systems, and model internals.

Mock Mate
Real-time AI mock interview platform with voice interaction and automated performance reports.
Next.jsReactTailwind CSSVapiGemini APIFirebase
  • Integrated Vapi WebRTC voice with Firebase Realtime Database for live interview sessions.
  • Built a Gemini evaluation pipeline with prompt chaining for structured performance reports.
  • Shipped secure auth and session storage with Firebase Authentication.
GitHubLive Demo
InfluenceIQ
AI-powered influencer ranking system across 950 Instagram profiles using sentiment and engagement signals.
PythonHugging FaceTransformersPandas
  • Ranked 950 profiles with RoBERTa sentiment analysis and log-scaled credibility metrics.
  • Designed a weighted scoring pipeline with dense ranking and Min-Max normalization.
  • Processed Instagram JSON across 15+ categories with multilingual comment classification.
GitHub
GPT from Scratch
End-to-end GPT implementation — attention, tokenization, training, and generation — built from first principles.
PythonPyTorchTransformersNLP
  • Implemented multi-head attention, GQA, KV-cache, and full transformer blocks from scratch.
  • Built BPE tokenization, data loaders, and a training loop with text generation.
  • Grounded the stack in custom neural-net primitives: backprop, MLP, activations, and loss.
GitHub

Skills

Languages

PythonJavaScriptGoC/C++

AI / ML

PyTorchHugging FaceOpenCVScikit-LearnNumPyPandasONNXTensorRTNVIDIA DeepStream

LLM Systems

LangChainLangGraphRAGChromaDBOllamaQdrant

Frontend

ReactNext.jsTailwind CSS

Backend

FastAPIExpress.jsNode.js

Databases

MongoDBMySQLFirebaseSQL

Tools

DockerGitLinux

Contact

Interested in building AI products or scalable backend systems? Let's connect.

EmailLinkedInGitHub