This study examines the poverty reduction implications of the introduction of three different agricultural technologies by government and NGOs in three rural sites across Bangladesh. The first is new vegetable seeds developed by AVRDC introduced in Saturia to women owning small amounts of land by a local NGO, based on a training and credit dissemination approach. The second is polyculture fish technology developed by WorldFish Center and introduced by a government extension program based on private fishponds operated mostly by men in Mymensingh. The third is the same polyculture fish technology, but introduced through a local NGO in Jessore based on the arrangement of leased fishponds operated by groups of low income women, supported by training and credit provision. The study found a number of significant poverty impacts. Among the strongest was in the case of the vegetable technology, which is targeted towards women in households with relatively small amounts of land and is a ‘non-lumpy' technology requiring a very low level of investment, but with proportionately significant returns and positive impacts on female empowerment and child nutritional status. The private fishpond technology was less successful in terms of poverty impact, since only better-off households tend to own ponds. This technology, however, had positive effects on the pond and crop profits of these households. The operation of the group fishpond technology, though a potentially beneficial agricultural program for poor households, was significantly undermined by collective action problems. Relative to women who did not have access to this group-based program, female group members appeared to have more mobility, greater likelihood of working for pay, higher off-farm incomes, and better nutritional status. The group fishpond technology was found also to increase vulnerability in a number of ways, such as through the theft of fish from ponds, or through gendered intra-household inequalities in (a) technology-related time burdens and (b) access to markets for and hence income from the agricultural outputs produced. The study overall showed a higher level of trust for NGO as opposed to government services, but also highlighted the variable performance of NGOs. Political dimensions to NGO activity also emerged as important, and are perceived by some sections of the community to affect the dissemination of technologies and extension support services for the technologies. Quantitative and qualitative data were found to complement each other well in the research across a range of issues. For example, the survey addressed female empowerment adopters by measuring the frequency of women's attendance of meetings etc, while the focus groups revealed the importance non-monetary exchange of vegetables between households to maintain social networks and reduce vulnerability. There were also gains through the overall use of the sustainable livelihoods framework as a way of sharpening understanding of the different entry points at which technology can impact on household well-being and vulnerability.