About the role
We're looking for a Data Scientist to own end-to-end ML models and experimentation across the company. This is a high-ownership, execution-heavy role focused on improving conversions, call success rates, and operational efficiency. You'll own problems end-to-end - from metrics and experiments to production models and product impact.
Responsibilities:
- Own end-to-end ML systems for an AI-driven outbound call centre.
- Work on problems directly tied to loan sales, collections, and insurance conversions.
- Drive experiments → product changes → measurable revenue impact.
- Build user / lead scoring models to decide who to call, when, and how.
- Improve conversion rates across outbound funnels (dial → connect → conversation → conversion).
- Work with rich, messy data: call transcripts, recordings, user metadata, and call-level signals.
- Extract insights from within-call behaviour (drop-offs, objections, engagement patterns).
- Design and run A/B experiments on model + product changes.
- Improve LLM agent performance (response quality, handling objections, conversation flow).
- Identify failure points (bad targeting, poor timing, weak conversations) and solve them using data + ML.
- Ship models into production and iterate based on real-world performance.
Requirements:
- Strong Python + SQL (used daily).
- Hands-on ML experience (practical projects/internships).
- Ability to go from raw data → insight → model → deployment.
- Comfort working with noisy, real-world datasets (text, behavioural data).
- Understanding of experimentation / A-B testing.
- Strong problem-solving + ownership mindset.
What Success Looks Like:
- You improve business metrics (conversion, revenue), not just model accuracy.
- Your work directly influences who gets called and how conversations are handled.
- You can break down a funnel and explain where and why it's failing.
- You independently pick problems and drive them end-to-end.
Tech Stack:
- Python, ML libraries (sklearn / PyTorch)
- ClickHouse
- LLMs (Claude, internal systems)
- Experimentation-driven environment