Harsh Kasat
Django Dev | Machine Learning | Freelancer | GenAI Dev | Software Engineer
About
A dev with chaotic thoughts and fluppy code, I learn by building product, breaking things is part of process. Mostly work remote because why not ? more into low-level and tinkering around system
Work Experience
KliqstrRemote (USA)
Gen Ai Developer (Freelancer)
Built AI agents to automate customer support and lead generation using LangChain + Superagent. Developed backend API with FastAPI to serve UGI models. Created voice modulator to clone user tones for video replies. Automated daily Facebook content across 30+ pages with Redis-powered pipelines — cut manual work by 90% .
OccultdiyRemote (London)
Django Developer (Contract Based)
Rebuilt core web app frontend for smoother UX and faster load times. Set up end-to-end CI/CD with GitHub Actions + Docker for zero-downtime deploys. Migrated backend from SQLite to Supabase, improving DB speed 3x. Sent targeted email campaigns to 15k+ users using cron + custom scripts.
AMD Telecom S.A.Remote (Greece)
Software Engineer
Designed and deployed 10+ microservices to handle Gen AI workloads using Docker + Redis. Built and fine-tuned invoice parsers that converted scanned invoices to structured JSON with 95% accuracy. Integrated NVIDIA RIVA to transcribe and synthesize voice calls in real time. Created custom image editing tools using Stable Diffusion + ControlNet for marketing automation.
Skills
Projects
A FastAPI application that leverages AI to generate emails, messages, and Upwork proposals. Key Features: Email Generation, Message Generation, Upwork Proposal Generation, Email Search, LLM Query
MeetAssistant is a Python-based automation tool that joins Google Meet meetings, records the session, transcribes the audio, and generates insightful reports based on the transcription. It leverages Selenium for browser automation, and integrates with AWS S3 for storing recordings and potentially other services for transcription analysis (LLM integration suggested but not fully implemented in provided code).
This project automates the creation of educational math animation videos using Manim, leveraging AI for script generation and video metadata creation. It integrates with Google Sheets for managing video ideas and uploaded video data.
This project implements an MCP server that provides tools for LLMs (Large Language Models) to interact with external data sources and execute operations to get best best MCP TOOL(Agent) Recommendation. It uses SSE (Server-Sent Events) as the transport layer, allowing for real-time communication between clients and the server.
Automated invoice generation system with PDF support.
AI-powered content generation tool for various writing tasks.
Clone of Bolt using Gemini API for AI functionalities.
AI tools for freight forwarders to automate shipments and provide real-time visibility.
Generates vector representations of user profiles for analysis.