I'm Rishikesh, an AI research engineer and technical founder working across model research and applied systems. I build multilingual speech synthesis, voice cloning and real-time conversational AI, alongside open-source implementations of language, audio and vision models.
I enjoy taking a paper apart, implementing its ideas in PyTorch, and exploring how to train and run models efficiently. My current interests include state-space and hybrid language models, GPU kernels, and coding agents with retrieval, persistent memory and tool use.
Leading multilingual dubbing, neural TTS, voice cloning, speech transformation and generative audio. Built agentic RAG dubbing workflows with human review, achieving approximately 2× faster dubbing. Developed multilingual foundation-model capabilities with RLHF-based post-training for translation quality and alignment.
Developed an in-house dubbing harness and tools, including an LLM-as-judge quality verifier, to automate dubbing and quality analysis, reduce reliance on human review, and save approximately 40% in time and 25% in costs.
—
AI Consultant / Developer · MyClone.is
Real-time voice AI, persona-aware retrieval, memory and conversational infrastructure. Remote, San Francisco-based company.
Oct 2018 — Jan 2019
Data Science / Software Development Engineer · Boutiqaat
Multimodal product matching, recommendations and catalog classification.
Dec 2017 — Oct 2018
Data Scientist · Humonics Global
Speech synthesis, question-answer generation, face recognition and liveness detection.
Jun 2015 — Dec 2017
Software Engineer · Nucleus Software
Banking analytics, loan-default prediction and enterprise financial software.
Taking models into real conversations and workflows.
AI consultant · Completed six-month engagement
MyClone
Developed voice and backend features for real-time digital personas: streaming conversations, persona-aware retrieval, document ingestion and persistent memory.