Harsh
Trivedi

Hey! I am a 3rd year Engineering student specializing in AI & Machine Learning. I like to work on Agentic Systems, RAG pipelines and study core ML/DL architectures.

B.E. AI & ML

NHCE Bengaluru

Graduating 2027

EmailLinkedInGitHubBlog
/01

Experience

TIFIN-India

AI Engineering Intern

Aug 2025 – Feb 2026

Hybrid · Bengaluru

  • →Contributed to production-grade agentic systems and built a Text-to-SQL tool enabling agent-to-database queries for funds, stocks, and AMC data (AUM) for financial advisors.
  • →Reduced container size of the entire application from 16 GB to 9 GB by creating a microservice that hosts embedding models like multilingual-mpnet-base-v2 instead of runtime in-memory loading.
  • →Worked on a document ingestion pipeline using MinerU extractors and evaluated varying chunking and embedding strategies for RAG-based retrieval.
  • →Worked on Schema migration across dev and production databases using Atlas schema migrator in Go
  • →Worked on a Web Search Module to allow the core agent to access Web Search via the Perplexity API

Plotch.ai

AI Intern

Apr 2025 – Aug 2025

Remote

  • →Built a pipeline to ingest unstructured data and generate rulebooks to derive SQL deviation queries using Gemini 2.5 Pro and Google AI Studio.
  • →Built Model Context Protocol (MCP) servers for GitHub workflows — view last merged PRs, get repo contents — using FastMCP by Anthropic.
  • →Implemented causal inference methods for multivariate data using the causal-learn library.
  • →Researched Reinforcement Learning and Multi-Agent RL; implemented environments and agents for RCA automation using OpenAI Gym and Stable Baselines 3.
/02

Projects

Oceo-Geo

March 2026

PythonFastAPIOllama, AgnoAI (agentic framework)
  • →OceoGeo solves the problem of using decade old plugins for analysing your data stored in NetCDF files by providing a detailed statistical engine, visualizations and an AI copilot so that you can talk to your data.
  • →Worked on the python backend including the AI service and the NetCDF parsing API .

Llama-3.2 3B Text-To-SQL

January 2026

LoRAUnslothGGUF
  • →Generated a schema specific natural language vs SQL Synthetic dataset for a schema speific Text to SQL fine tuning using LoRA
  • →Coverted model to GGUF and used it for local inference via Ollama

GitHub & Obsidian MCP

June 2025

PythonFastMCPGitHub API
  • →Developed a Model Context Protocol server from scratch using GitHub APIs and Obsidian functions with Claude as the client.
  • →Obsidian integration lets you automate creating, reading, appending, and deleting notes; GitHub server surfaces repository contents, file content, and last-merged PR details.
  • →Enables tasks like writing notes in Obsidian after Claude analyses your repositories.

HydRAG

April 2024

PythonLangChainHuggingFaceFAISS
  • →Implemented Retrieval-Augmented Generation for groundwater data using the open-source Google FLAN-T5 and Google Tapas models via HuggingFace.
  • →Used LangChain for the RAG framework, HuggingFace for model access, and FAISS as the vector database.
/03

Skills

Core Focus

PythonFlaskPostgreSQLDockerAI AgentsLangChainPhidata

Languages

JavaCJavaScriptHTML / CSS

Frameworks & Libraries

FastAPIFastMCPNext.jsPyTorchGym (OpenAI)

Tools & Platforms

GitGitHubGitLabMySQL
/04

Education

New Horizon College of Engineering

Bachelor of Engineering — Artificial Intelligence & Machine Learning

Bengaluru, Karnataka

Oct 2023 – Jun 2027

9.04CGPA
/05

Achievements

2025

National Hackathon Winner

Quantum x 2025

Secured first place at a national-level hackathon for developing an innovative AI-driven solution.

2025

Seminar Host & Speaker

New Horizon College of Engineering

Conducted a comprehensive 2-day seminar on AI Agents and Model Context Protocol (MCP) for the faculty of the AIML department.