Rachit Mukherjee

AI Video Subtitle Generation Platform (Internal AI Tool): Developed an end-to-end automated subtitle generation system using Whisper, LLaMA 3 (Ollama), FFmpeg, and custom DSPy pipelines. Designed a fully automated workflow for transcription → translation → timestamp alignment → subtitle formatting (.srt/.vtt). Built a Next.js internal dashboard enabling batch-processing and review; increased video production efficiency by 45–60%. Implemented multi-language support (Hindi/English) with post-processing correction using LLaMA3. Optimized FFmpeg-based video chunking to reduce processing time by 35%.

MOAI – Multi Modular Agentic AI: Built modular multi-agent architecture using Ollama & Hugging Face models for conversational AI, RAG (Weaviate), and multimodal generation. Reduced inference latency by 30% through prompt-routing optimization and model caching. Designed scalable unified interface increasing team productivity by 40%. Achira Mobile Application (Flutter): Developed cross-platform news app with AI summarization & sentiment analysis; reduced reading time by 60–80%. Published on Play Store & App Store, reaching 1k+ active users. News Summarisation Model – Achira News: Built custom LLM-based summarizer reducing information overload by 65%. Automated editorial workflow resulting in 50% faster news processing. Ginger Bell xAPI Backend Service: Engineered Django backend with OpenRemote integration; improved API reliability by 25%. Implemented RBAC and secure token workflows reducing auth errors by 40%. MQTT OSS System (Hybrid Cloud Architecture): Designed IoT cloud infrastructure with OpenRemote, Grafana, Prometheus, Nginx, Jenkins, Docker. Improved system uptime by 20% and reduced deployment overhead by 30%. AIoT-Edge Smart Doorbell & Security Camera: Built ESP32 + Raspberry

Key Coursework: AI, ML, Deep Learning, DSA, Web Tech, DBMS, Blockchain, DevOps