Research Engineer

About Dashverse

Dashverse is building the next-gen platform for AI-native visual storytelling. Three platforms — Dashtoon (comics), Dashreels (vertical drama), and Studio (creator tools) — let creators around the world make and monetise high-quality content at scale. 100K+ MAUs, creators in 20+ countries, backed by world-class investors. Founders have built at Udaan, PocketFM, and more.

The Role

This is a research seat, not a forward-deployed engineering one. You'll work close to the frontier of generative AI — diffusion models, character consistency, video generation, multimodal systems — and ship real models to real users.

We're building proprietary models for proprietary problems. Not chasing open-source releases — staying just behind frontier labs, reading their research, prototyping quickly, productionising what works.

Must Haves

• 3-8 years of ML / research engineering experience
• Strong fundamentals: PyTorch, model training, optimisation, evaluation
• Track record of building or training real models — not just API integration
• Comfort reading recent papers and reproducing them
• Production sensibility — you can take a model from notebook to scaled inference

Nice to Have

• Publications at top venues (CVPR, NeurIPS, ICML, WACV, ICLR, ACL, EMNLP, etc.)
• Patents or notable open-source contributions
• Diffusion model experience (image, video, or audio)
• Multimodal / vision-language model experience
• GPU optimisation, inference engineering, or training infrastructure
• Experience with character consistency, video generation, or AI content pipelines
• Background in computer vision, NLP, or speech

What You'll Work On

• Push the frontier on visual storytelling models — character consistency, scene composition, video generation
• Read frontier research, prototype quickly, productionise what works
• Own the full model lifecycle: data, architecture, training loop, evaluation, deployment
• Collaborate with product and engineering on what to build next
• Help shape the research agenda for visual storytelling

What We Offer

• Real research environment with serious compute (H100/H200)
• Frontier-adjacent work, building proprietary models for proprietary problems
• Well-funded, growing fast — no near-term existential pressure
• Bangalore office (remote considered for exceptional candidates)
• 50-80 LPA based on experience and depth