About the role
Head of AI, Vertical AI Agents at Tamasha.live - India's Most Exciting Live Entertainment Platform.
The Role:
You'll own the AI platform that every Tamasha agent runs on, and the business results those agents produce. The job is to make each agent predictable: it stays in character, remembers what matters, gets its domain right, and is safe and compliant, every time. This is a high-visibility role: you will report directly to the Co-Founder and work across Tamasha's multi-app ecosystem.
What You'll Do:
- Build the shared platform: a reusable agent framework, long-term memory, evaluations, guardrails and domain tools.
- Set the model strategy. Choose and combine open-source LLMs (Llama, Qwen, Gemma, Indic models), using hosted APIs only where they're worth it. Fine-tune (SFT, DPO, LoRA) when it improves quality or cuts cost.
- Own unit economics. Lower the cost per conversation with routing from small to large models, distillation, caching and quantised self-hosting.
- Make every agent predictable: persona consistency, memory design, accuracy, and safety on sensitive topics.
- Build evaluations, offline and online, including red-teaming, tied to conversation-to-transaction conversion and to retention.
- Work with domain experts to turn their knowledge into deterministic tools and grounded content.
- Protect user data: comply with the DPDP Act, and give users control over what agents remember.
- Hire, mentor and grow the AI team, starting with a Staff AI Engineer.
What You'll Need:
- Hands-on approach.
- Around 10 years or more in ML or NLP, including shipping LLM or conversational systems to real users at scale.
- Hands-on depth in fine-tuning and evaluating open-source models, and in retrieval and memory design.
- A track record of cutting inference cost and latency without losing quality.
- Experience making model behaviour measurable and reliable through evaluations, guardrails and regression checks.
- Business judgement: you choose models and features by their effect on conversion, retention and margin.
- Experience leading engineers and explaining trade-offs clearly to founders.
Requirements
- Hands-on approach.
- Around 10 years or more in ML or NLP, including shipping LLM or conversational systems to real users at scale.
- Hands-on depth in fine-tuning and evaluating open-source models, and in retrieval and memory design.