Nadeem Akhtar
AI Researcher & Engineer · Bengaluru, India
I work on efficient large language models: parameter-efficient fine-tuning, federated learning, compression, and reasoning evaluation. At Synaptics, I build AI systems that run in production.
Selected results
53.1%
less communication in federated LoRA
With lower final loss, across 7 seeds on Llama-3.2-3B and Qwen2.5-3B.
Best AI Work
Synaptics AI Day
AP Invoice Automation: live for the Korea division; rollout underway across 8 global regions.
90.3 pts
reasoning-elicitation collapse
A 90.3-point reasoning-elicitation collapse, from English (91.6%) to Hindi (1.3%).
Where to start
- For research labs and PhD supervisors Research 2 accepted · 2 under review · 1 in preparation Publications, research threads, and public models and datasets on Hugging Face. See research
- For engineering teams Engineering Three AI systems I built at Synaptics, each with the problem, approach, result, and stack. See case studies
- For clients Freelance Remote contract work for US teams: document automation, LLM apps and agents, fine-tuning, and evaluation. See services