Voice and chat agent testing and security
Hamming AI
Hamming AI builds a testing, red-teaming and monitoring platform for voice and chat agents. Its users include engineering, product and quality-assurance teams deploying conversational AI in banking, healthcare and customer support. Teams use it to check whether agents complete tasks correctly, follow required policies and protect sensitive information, both before release and during live customer conversations.
Hamming tests the conversation as audio, including failures that a transcript can obscure. Simulated callers introduce accents, background noise and interruptions; turn-by-turn diagnostics capture timing and transcription and separate delays across speech recognition, language models and speech synthesis. Its testing and monitoring workflow connects those investigations to repeatable checks: a failure discovered in production can become a regression test that runs on subsequent releases.
Founded by Sumanyu Sharma, Hamming began with general AI evaluation infrastructure before shifting to voice-agent testing and launching that product in August 2024. It subsequently extended the same call-simulation machinery into adversarial testing, probing how agents respond to malicious callers. Today, its security work includes prompt injection, jailbreaks, social engineering and attempts to extract personal information. Hamming operates independently of agent platforms, model providers and telephony vendors, allowing teams to test across the components they choose.
1 talk
Newest firstLoading more talks and speakers…
1 speaker at AIE
Affiliations reflect their AIE event listings and recorded appearances, not necessarily current employment.
