Research
Select a project for the problem, my contribution, method and results.
Language models & agents
- Small LM Post-Training for AI Call Agents Synthetic training data and one LoRA recipe for six small LMs that answer calls. My roleData pipeline, LoRA fine-tuning and evaluation.
- CallScreenBench A benchmark for small LMs that answer calls on their owner's behalf. My roleFirst author (equal contribution); built the benchmark end to end.
- Early Risk Prediction in Real Calls Predicting requests for sensitive information from a call's first lines. My rolePrediction analysis on 10,211 real calls.