We study the impact of air pollution exposure (CO, O3 and Pm10) during pregnancy and early years of life on infant health for a sample of children attending public kindergartens in Bogotá, Colombia. The study uses a unique database that gathers information on children health which allows to combine information of residential location of the mother with information from the city air quality monitors. To overcome endogeneity problems due to residential sorting we identify pairs of siblings in the dataset and implement panel data models with mother fixed effects. Results show evidence that mothers, who are exposed to higher levels of CO and O3 during pregnancy, have a higher probability of their babies being born with a low birth weight. Furthermore, a child exposed in-utero to higher levels of O3 has a higher probability of being diagnosed with a lung-related disease. Our findings advocate for more strict environmental regulations as a way to improve human capital in developing countries.