I’m Mitra, an ML engineer at Lucid Axon! We are a Fintech startup working to improve investing by analyzing company’s performance and ESG alignment. My work is particularly focused on large language models, text feature extraction, and building chat models!

🔗 GitHub

Paper reading journey

Notes on Statistics

🍀Some of my personal projects

<aside> 🤖 Generating QA for fine-tuning Chat

Tools: Supervised Fine-tuning, LangChain, Huggingface🤗, LoRA

🔗 GitHub Repository

</aside>

<aside> 🤖 RAG on PDF files

Tools: PostgreSQL, LangChain, HuggingFace Embedding

🔗 GitHub Repository

</aside>

<aside> 🤖 Agentic AI Containerized app for summarizing Wikipedia

Tools: Docker, LangChain, PostgreSQL

🔗 GitHub Repository

</aside>

<aside> 🤖 Conversation analysis on Social Media

Tools: Zero-shot & Few-shot learning

🔗 GitHub Repository

</aside>

<aside> 🤖 Web scraper and summarizer - Traditional modularized package

Containerized app for summarizing web content

Tools: Docker, LangChain, PostgreSQL

🔗 GitHub Repository

</aside>

<aside> 🤖 PassivePy

A Python package for identifying passive voice in text data. Built with spaCy and regex patterns.

🔗 GitHub Repository

</aside>

<aside> 📱 COPER Search Engine

A semantics-based search engine for Persian COVID-19 articles using BERT and elastic search.

🔗 GitHub Repository

</aside>

latest courses/certificates

<aside> 🤖 AI agents

AI Agents in LangGraph

Tools: LangChain

🔗 Certificate

</aside>

<aside> 🤖 RAG (Langraph + NoSQL)

Tools: Langraph

🔗 Certificate

</aside>

<aside> 📝 Prose2Poem

A transformer-based model for generating Persian prose.

🔗 GitHub Repository

</aside>

<aside> 📊 Social Media Analysis Pipeline

An end-to-end pipeline for analyzing social media data for municipal governance using LLMs.

🔗 GitHub Repository

</aside>

📜Resume

Experience

NLP Engineer – Lucid Axon, Montreal, CA (Aug. 2024 - Present)

FinTech Startup - Project-based fulltime - Remote

Prompt engineering, hyperparameter tuning, and fine-tuning Meta-LLaMA, Mistral AI, and DeepSeek models on chunked legal and financial documents for an ESG-aware chatbot
Developed an end-to-end feature classification pipeline using a hybrid of two LLaMA models and statistical models, achieving an F1 score above 85% across all classes
Designed a retrieval-augmented generation (RAG) system integrating LangGraph and FlashRank with a NoSQL database
Managing documentation and progress tracking via Jira and Bitbucket for Agile development

ML Engineer - School of Public Policy, Calgary, CA (Oct. 2022 - Present)

Public Policy research group – Part-time contract – Remote

Built and optimized sentiment classification and topic modeling pipelines using LLMs for large-scale social media analysis
Created interpretability dashboards and visualizations using D3.js and Plotly