Divanshu Sharma — Applied AI / ML Systems Engineer

run 00 · cached

“whoami”

▸ intent: about
▸ tool: get_about()
▸ synthesis: cached

Divanshu Sharma is the Founding Engineer at Uniiq.ai, building reliable AI products across React, Express, MongoDB and Gemini. His work combines stateful admissions workflows, model failure recovery and safe data updates. He also designed and validated an Admissions Decision Twin in controlled offline evaluation. Start with the hiring brief for ownership and evidence, or explore a system in detail.

Divanshu Sharma

Applied AI / ML Systems EngineerLLM evaluation, inference & reliability

Founding Engineer @ Uniiq.ai · Mumbai, India

Open to AI engineering & founding roles

Founding engineer who ships reliable AI products. At Uniiq he owns the AI admissions platform: a resumable, stateful intake with validated model outputs and failure recovery, an 80.8% smaller frontend entry bundle, and a Decision Twin designed and validated in controlled offline evaluation. Before that: privacy-preserving deep learning research at Yale and three years of quantitative research at WorldQuant BRAIN.

80.8%
smaller frontend entry bundle
Uniiq · 1.45 MB → 278 KB · 81 KB gzipped
28
merged upstream PRs
Mastra · EleutherAI lm-eval · 40 issue reports also resolved upstream
88.08%
ChestMNIST accuracy under MPC
from-scratch DL library, Yale
70% → 0%
critical-constraint violations
Uniiq Decision Twin · offline evaluation

Uniiq.ai · Admissions Decision Twin — evaluation & results →