Together with a very talented team, we have been building ELSA (English Language Speech Assistant) with a big vision to enable everyone in the world to speak foreign languages with full confidence, reaching the best life and career opportunities. Our flagship product, ELSA Speak, is a personal tutor which helps users improve their spoken English communication skills through extensive use of AI.As a founder I have worn many hats over the years, here is a shallow summary:## 2015 to 2020 - first 5 years* I built the research, engineering and data teams from scratch. From day one we have grown hybrid teams, with offices in Lisbon, Portugal, India and Vietnam.* In engineering, we developed 2 mobile apps (native Android and IOS) and a micro-services backend architecture.* Around 2017 we containerized and introduced Kubernetes, deciding on it over docker swarm (popular at the time).* In research we first focused on improving pronunciation assessment and feedback and later added fluency and prosody (sentence stress) analysis.* Around 2018 we introduced the assessment test as a way for the user to get an initial placement, tracking of all the above scores over time and visualization via a Pentagon.* We also introduced our first recommender engine, focused on user skill tracking in real time, compared via an ELO algorithm to a crowdsourced and real time estimation of the difficulty of each lesson.* In late 2019 and early 2020 we started working with end2end ASR for highly accented non-native speech.## 2021 -2022* As the company kept growing bigger, we decided to hire a Senior VP of Engineering to own engineering and business intelligence, allowing me to focus on AI research.* Our ASR accuracy continued to increase, which allowed us to implement technology for grammar and vocabulary analysis, leading to our Speech Analyzer web product.## 2023* Thanks to the general availability of chat-tuned LLM models we started developing our ELSA Tutor vision

I am helping Vilynx change the way we watch videos. I am involved as a external advisor for the areas of speech and audio analytics.In October 2020 it has been announced Vilynx has been acquired by Apple.

Sinkronigo was born after realizing how technology could help the quickly emerging digital publishing market. With a strong background in speech processing, we specialize in creating read aloud epub3 eBooks. Our books contain human narrations that are aligned with the text up to word level, allowing the reader to easily follow what is being said as the text is highlighted as it is being spoken.

participated in the Startx acceleration program as co-founder of Elsa

I helped create a new Toastmasters club within Telefonica Barcelona. We chartered the club in December 2015.

Research in multimedia (Audio+video+image+text) indexing. Developing tools to help users search and retrieve information either through TV sets, computers or cellphones.

Speech processing research and development, focusing on acoustic indexing.

I taught courses on programming, operating systems and digital speech processing. In most cases I was appointed as the official professor for the given course, which meant that I needed to define the curricula, manage the different people that helped me in lab and practical classes, prepared exams and performed the final evaluations.

Prestigious Speakers Toastmasters club president in Barcelona, Spain.

I pursued a research stay/post-doc position for almost 2 months, working on the improvement of the Spanish speech recognition system to apply to the TC-Star project for European Parliament translation.

I'm working on speaker segmentation and clustering for meeting room environments. The first year I received a grant from the AMI Training Programme (6th EU framework programme). The second year I received a grant from the Spanish Education Ministry.

Area E5 in District 57, Downtown Berkeley.Led 4 clubs into achieving their goals and reported to the District governor in the monthly meetings.Achieved President's distinguished Area during my term.

I developed algorithms to perform speaker segmentation (finding the temporal boundaries between speakers) and later speaker diarization (finding speaker boundaries and group together segments spoken by each speaker.

I developed the Spanish Panasonic TTS (text-to-speech) engine from end to end, including text selection, prompts recording, segmentation and forced/manual alignment and adaptation of an inhouse TTS engine to Spanish. Later I helped improve prosody for several other languages and worked on language-independent TTS

I was an intern at the large format printers division, helping on the research and development of the writting system algorithms.