Where we’ll be next
Heading to the same event? We’d love to meet you.
Where we’ll be next
Heading to the same event? We’d love to meet you.
Where we’ll be next
Heading to the same event? We’d love to meet you.
24
OCT
EMNLP 2026
Recently Attended
24
OCT
EMNLP 2026
Recently Attended
-
Check back soon
No scheduled events
-
Check back soon
No scheduled events
Past events
A look back at the events and conferences we’ve attended, the conversations we’ve had and what we’ve learned along the way.
A look back at the events and conferences we’ve attended, the conversations we’ve had and what we’ve learned along the way.
EMNLP 2026
·
Budapest, Hungary
·



Our first EMNLP was a great few days spent meeting researchers and engineers from across the natural language processing (NLP) and AI research community. It was a chance to catch up with familiar faces, meet plenty of new people and get a closer look at some of the research shaping the field.
EMNLP is one of the leading conferences for NLP and language modelling, and the scale of this year's event was impressive. With more than 5,200 papers accepted across the main conference and Findings, there was a huge amount of research to explore, covering large language models, AI agents, multimodal learning and plenty more.
One of the standout themes was agentic AI, particularly how language models use tools, manage memory and tackle more complex tasks. There was also interesting work around reinforcement learning, LLM reasoning, post-training and evaluation, with a recurring focus on making AI systems more reliable beyond controlled benchmarks.
It was interesting to see how much NLP now overlaps with other areas of AI research. Alongside language modelling and multilingual NLP, the programme included work on speech and audio, vision-language models, robotics and embodied AI. The range of topics made for some great conversations about where the research is heading.
As always, one of the best parts was speaking directly with the people behind the research. Chris Farish and Kieran Newton were getting stuck into conversations throughout the week, reconnecting with familiar faces and meeting researchers for the first time. There were plenty of interesting discussions, alongside a few comments about the number of AI-generated presentation slides!
Away from the conference, we managed to explore a bit of the city, enjoy some excellent food and fit in a few games of pool. The venue did a brilliant job handling the crowds, and it was a really enjoyable few days all round.
A big thank you to everyone we caught up with and to the organisers for putting on such a great event. It was brilliant to spend time with the NLP research community, hear about the work happening behind the scenes and come away with plenty of new connections and conversations to continue.
Key Takeaways:
Research we saw:
Hiring trends:
ICASSP 2026
·
Barcelona, Spain
·



ICASSP 2026 was a particularly special conference for Techire AI. After seven years specialising in Speech & Audio recruitment, we were proud to sponsor and exhibit at one of the world's leading conferences in acoustics, speech and signal processing, becoming its first dedicated recruitment partner to do so.
With almost our entire team attending, we spent the week meeting researchers, engineers and hiring managers from across the speech AI community. Our stand was busy throughout the conference, with plenty of familiar faces stopping by alongside people we were meeting for the first time.
The research programme covered a huge range of topics, from speech recognition and audio generation to machine learning, signal processing and computer vision. Naturally, speech and audio AI were our main areas of interest, particularly the developments in real-time conversational AI and the challenges of building more natural, responsive voice systems.
One of the standout topics was full-duplex speech, where models can listen and respond simultaneously rather than relying on traditional turn-taking. There was also interesting work around automatic speech recognition (ASR), text-to-speech (TTS), speech enhancement and audio-language models. These are particularly relevant to the teams we work with developing conversational AI and voice technology.
On-device AI was another interesting area of discussion, particularly how speech models can run efficiently on local hardware without compromising performance. Model compression, latency and computational efficiency are increasingly important considerations as speech technology moves into everyday products. We were also surprised by the sheer amount of computer vision research, including image processing, video understanding and multimodal AI.
Beyond the research, one of the biggest highlights was reconnecting with people we've known for years, including researchers we've helped into speech AI roles dating back to 2020. We also met candidates we'd been working with remotely, caught up with clients and even managed to close a placement with a candidate visiting from San Francisco.
Away from the conference floor, it was a rare opportunity to get most of our remote team together in one place. We managed to explore a bit of the city, fit in some indoor skydiving and enjoy a few beers along the way.
A huge thank you to everyone who visited our stand, shared their research, entered our giveaway or simply came over to say hello. Being part of a community we've worked with for so long made this one of our most memorable conferences yet.
Key Takeaways:
Research we saw:
Hiring trends:
NeurIPS 2025
·
San Diego, California
·



Our first NeurIPS, and we completely underestimated the scale of it.
Marc and Chris headed to San Diego for three days of research talks, poster sessions, meetings and side events at one of the biggest conferences in AI. We've been to plenty of conferences over the years, but NeurIPS was something else. The sheer number of people, papers and events made it difficult to know where to start.
The range of research was impressive, covering everything from LLM reasoning and reinforcement learning to multimodal AI, computer vision, robotics, speech and audio, and neuroscience. There was so much happening across the conference that deciding what to attend was a challenge in itself.
A lot of the research caught our attention, particularly around reasoning models, agentic AI and how reinforcement learning is being used to improve model capabilities. Papers like The Illusion of Thinking and Why Do Multi-Agent LLM Systems Fail? raised interesting questions about what these systems can actually do reliably, beyond performing well on benchmarks.
There was also plenty of work on making models more efficient, with research into speculative decoding, KV cache compression and quantisation. It's an area that's becoming increasingly relevant as more teams look at the practicalities of deploying large models rather than just improving their performance.
Multimodal research was another big part of the programme. From vision-language models and video generation to 3D reconstruction, robotics and audio-language models, it was interesting to see just how much overlap there is between areas that were previously more separate. Papers such as PerceptionLM and Audio Flamingo 3 were good examples of the breadth of work being presented.
One of the highlights was finally meeting researchers and industry contacts we'd known for years but hadn't had the chance to meet in person. We also caught up with familiar faces, made new connections and had some great conversations with people working across different areas of AI.
NeurIPS had a different feel from smaller, more specialised conferences we've attended, with a much broader mix of researchers, founders, engineering leaders and investors.
Beyond the main conference, we explored the exhibition hall, attended the Hugging Science party and made the most of San Diego's sunshine.
A brilliant first experience, and definitely one we'll be returning to.
Key Takeaways:
Research we saw:
Hiring trends:
ACL 2025
·
Vienna, Austria
·



Marc and Chris attended ACL 2025 in Vienna, joining researchers from across the world for one of the leading conferences in computational linguistics and natural language processing.
With so much research being presented, two themes particularly stood out to us: model architecture and data. Both came up repeatedly in conversations throughout the conference, particularly around how large language models (LLMs) are trained, how they're evaluated and where researchers are looking to make improvements.
There was some interesting work exploring alternatives to the way language models are currently built, including tokenisation, attention mechanisms and more efficient approaches to training and inference. With so much focus on scaling models, it was good to see research looking more closely at the architecture itself and whether there are better ways of doing things.
Data was the other big topic for us. From pretraining and fine-tuning to reinforcement learning and preference optimisation, there were plenty of different approaches being explored. It also raised some interesting questions around data quality, how models learn and how much their behaviour can be influenced by the way they're trained.
The wider research programme covered everything from multilingual NLP and machine translation to reasoning, multimodal learning, AI agents and model alignment. There was certainly no shortage of interesting papers, and with so many sessions running alongside each other, it was impossible to see everything.
Beyond the talks and research presentations, it was a great opportunity to reconnect with researchers we've got to know over the years, meet new people working across the field and catch up with a couple of our clients in person.
Those conversations are always one of the most valuable parts of attending conferences like ACL. It's one thing reading someone's research or speaking over a video call, but quite another getting to sit down and hear what they're working on, what they're finding difficult and where they see things going next.
It was a great few days in Vienna, with plenty of interesting research to follow up on and some really good conversations with people across the NLP community.
Thanks to everyone who took the time to meet us. We're looking forward to the next one.
Key Takeaways:
Research we saw:
Hiring trends:
EMNLP 2026
·
Budapest, Hungary
·



Our first EMNLP was a great few days spent meeting researchers and engineers from across the natural language processing (NLP) and AI research community. It was a chance to catch up with familiar faces, meet plenty of new people and get a closer look at some of the research shaping the field.
EMNLP is one of the leading conferences for NLP and language modelling, and the scale of this year's event was impressive. With more than 5,200 papers accepted across the main conference and Findings, there was a huge amount of research to explore, covering large language models, AI agents, multimodal learning and plenty more.
One of the standout themes was agentic AI, particularly how language models use tools, manage memory and tackle more complex tasks. There was also interesting work around reinforcement learning, LLM reasoning, post-training and evaluation, with a recurring focus on making AI systems more reliable beyond controlled benchmarks.
It was interesting to see how much NLP now overlaps with other areas of AI research. Alongside language modelling and multilingual NLP, the programme included work on speech and audio, vision-language models, robotics and embodied AI. The range of topics made for some great conversations about where the research is heading.
As always, one of the best parts was speaking directly with the people behind the research. Chris Farish and Kieran Newton were getting stuck into conversations throughout the week, reconnecting with familiar faces and meeting researchers for the first time. There were plenty of interesting discussions, alongside a few comments about the number of AI-generated presentation slides!
Away from the conference, we managed to explore a bit of the city, enjoy some excellent food and fit in a few games of pool. The venue did a brilliant job handling the crowds, and it was a really enjoyable few days all round.
A big thank you to everyone we caught up with and to the organisers for putting on such a great event. It was brilliant to spend time with the NLP research community, hear about the work happening behind the scenes and come away with plenty of new connections and conversations to continue.
Key Takeaways:
Research we saw:
Hiring trends:
ICASSP 2026
·
Barcelona, Spain
·



ICASSP 2026 was a particularly special conference for Techire AI. After seven years specialising in Speech & Audio recruitment, we were proud to sponsor and exhibit at one of the world's leading conferences in acoustics, speech and signal processing, becoming its first dedicated recruitment partner to do so.
With almost our entire team attending, we spent the week meeting researchers, engineers and hiring managers from across the speech AI community. Our stand was busy throughout the conference, with plenty of familiar faces stopping by alongside people we were meeting for the first time.
The research programme covered a huge range of topics, from speech recognition and audio generation to machine learning, signal processing and computer vision. Naturally, speech and audio AI were our main areas of interest, particularly the developments in real-time conversational AI and the challenges of building more natural, responsive voice systems.
One of the standout topics was full-duplex speech, where models can listen and respond simultaneously rather than relying on traditional turn-taking. There was also interesting work around automatic speech recognition (ASR), text-to-speech (TTS), speech enhancement and audio-language models. These are particularly relevant to the teams we work with developing conversational AI and voice technology.
On-device AI was another interesting area of discussion, particularly how speech models can run efficiently on local hardware without compromising performance. Model compression, latency and computational efficiency are increasingly important considerations as speech technology moves into everyday products. We were also surprised by the sheer amount of computer vision research, including image processing, video understanding and multimodal AI.
Beyond the research, one of the biggest highlights was reconnecting with people we've known for years, including researchers we've helped into speech AI roles dating back to 2020. We also met candidates we'd been working with remotely, caught up with clients and even managed to close a placement with a candidate visiting from San Francisco.
Away from the conference floor, it was a rare opportunity to get most of our remote team together in one place. We managed to explore a bit of the city, fit in some indoor skydiving and enjoy a few beers along the way.
A huge thank you to everyone who visited our stand, shared their research, entered our giveaway or simply came over to say hello. Being part of a community we've worked with for so long made this one of our most memorable conferences yet.
Key Takeaways:
Research we saw:
Hiring trends:
NeurIPS 2025
·
San Diego, California
·



Our first NeurIPS, and we completely underestimated the scale of it.
Marc and Chris headed to San Diego for three days of research talks, poster sessions, meetings and side events at one of the biggest conferences in AI. We've been to plenty of conferences over the years, but NeurIPS was something else. The sheer number of people, papers and events made it difficult to know where to start.
The range of research was impressive, covering everything from LLM reasoning and reinforcement learning to multimodal AI, computer vision, robotics, speech and audio, and neuroscience. There was so much happening across the conference that deciding what to attend was a challenge in itself.
A lot of the research caught our attention, particularly around reasoning models, agentic AI and how reinforcement learning is being used to improve model capabilities. Papers like The Illusion of Thinking and Why Do Multi-Agent LLM Systems Fail? raised interesting questions about what these systems can actually do reliably, beyond performing well on benchmarks.
There was also plenty of work on making models more efficient, with research into speculative decoding, KV cache compression and quantisation. It's an area that's becoming increasingly relevant as more teams look at the practicalities of deploying large models rather than just improving their performance.
Multimodal research was another big part of the programme. From vision-language models and video generation to 3D reconstruction, robotics and audio-language models, it was interesting to see just how much overlap there is between areas that were previously more separate. Papers such as PerceptionLM and Audio Flamingo 3 were good examples of the breadth of work being presented.
One of the highlights was finally meeting researchers and industry contacts we'd known for years but hadn't had the chance to meet in person. We also caught up with familiar faces, made new connections and had some great conversations with people working across different areas of AI.
NeurIPS had a different feel from smaller, more specialised conferences we've attended, with a much broader mix of researchers, founders, engineering leaders and investors.
Beyond the main conference, we explored the exhibition hall, attended the Hugging Science party and made the most of San Diego's sunshine.
A brilliant first experience, and definitely one we'll be returning to.
Key Takeaways:
Research we saw:
Hiring trends:
ACL 2025
·
Vienna, Austria
·



Marc and Chris attended ACL 2025 in Vienna, joining researchers from across the world for one of the leading conferences in computational linguistics and natural language processing.
With so much research being presented, two themes particularly stood out to us: model architecture and data. Both came up repeatedly in conversations throughout the conference, particularly around how large language models (LLMs) are trained, how they're evaluated and where researchers are looking to make improvements.
There was some interesting work exploring alternatives to the way language models are currently built, including tokenisation, attention mechanisms and more efficient approaches to training and inference. With so much focus on scaling models, it was good to see research looking more closely at the architecture itself and whether there are better ways of doing things.
Data was the other big topic for us. From pretraining and fine-tuning to reinforcement learning and preference optimisation, there were plenty of different approaches being explored. It also raised some interesting questions around data quality, how models learn and how much their behaviour can be influenced by the way they're trained.
The wider research programme covered everything from multilingual NLP and machine translation to reasoning, multimodal learning, AI agents and model alignment. There was certainly no shortage of interesting papers, and with so many sessions running alongside each other, it was impossible to see everything.
Beyond the talks and research presentations, it was a great opportunity to reconnect with researchers we've got to know over the years, meet new people working across the field and catch up with a couple of our clients in person.
Those conversations are always one of the most valuable parts of attending conferences like ACL. It's one thing reading someone's research or speaking over a video call, but quite another getting to sit down and hear what they're working on, what they're finding difficult and where they see things going next.
It was a great few days in Vienna, with plenty of interesting research to follow up on and some really good conversations with people across the NLP community.
Thanks to everyone who took the time to meet us. We're looking forward to the next one.
Key Takeaways:
Research we saw:
Hiring trends:

