About
Articles by David
-
AI, what can be next?
AI, what can be next?
You may be aware already about the latest in AI with AlphaGo Zero, it is hard not to be impressed, but what can be…
1 Comment -
Data Reply UK is hiring
Data Reply UK is hiring
We are hiring! @DataReplyUK #SMR10 https://twitter.com/milkroundabout/status/660383478408589312
Activity
-
Today we released cross-cloud Lakehouse for Apache Iceberg to query data from an AWS Glue catalog directly within Google Cloud ! Storage federation…
Today we released cross-cloud Lakehouse for Apache Iceberg to query data from an AWS Glue catalog directly within Google Cloud ! Storage federation…
Liked by David Sabater Dinter
-
Low-key the most important product manager superpower? Talking to customers. Not talking at customers — PM-splaining the feature you're already sold…
Low-key the most important product manager superpower? Talking to customers. Not talking at customers — PM-splaining the feature you're already sold…
Liked by David Sabater Dinter
-
Conversational Analytics in BigQuery is now in GA - we have deeply integrated the agents with advanced BigQuery capabilities (BigQuery ML, graph…
Conversational Analytics in BigQuery is now in GA - we have deeply integrated the agents with advanced BigQuery capabilities (BigQuery ML, graph…
Liked by David Sabater Dinter
Experience & Education
-
Google
View David’s full experience
See their title, tenure and more.
or
Licenses & Certifications
-
-
-
-
-
-
DB2 9.1
IBM
-
-
Technology Architect
Accenture
Volunteer Experience
-
Volunteer
Aldeas Infantiles SOS de España
- 1 year 1 month
Children
I was volunteer at Aldeas Infantiles for one year over the weekends in one of the houses, helping organising events with the children and teaching IT skills to the teenagers.
Projects
-
Parallel Interference Cancellation in beyond 3G multi-user and multi-antenna OFDM systems
-
In the present thesis, the concept for beyond 3G mobile radio systems is described. A service area concept is introduced in order to combat the performance limiting interferences present in the cellular mobile communication systems, with each service area consisting of a various simultaneously active mobile terminals, a number of fixed access points and a central unit performing signal processing. About uplink transmission, the main characteristic of this service area system is that with the…
In the present thesis, the concept for beyond 3G mobile radio systems is described. A service area concept is introduced in order to combat the performance limiting interferences present in the cellular mobile communication systems, with each service area consisting of a various simultaneously active mobile terminals, a number of fixed access points and a central unit performing signal processing. About uplink transmission, the main characteristic of this service area system is that with the aid of joint detection of the transmit signals from the mobile ter- minals performed at uplink transmission, all interferences between the simultaneously active mobile terminals using the same bandwidth is drastically reduced. Moreover the use of OFDM subcarrierwise in the described service area based system allows for intersymbol interference free communication and for simple equalization in the frequency domain.
Through this subcarrierwise equalization the service area based system is equivalent with a number of smaller parallel systems, a fact that affects in a reduced computational complexity in the case of optimum multiuser operating with the maximum likehood principle and suboptimum linear detector zero-forcing. In this thesis the parallel interference cancellation detector is introduced, according to which the multi access interference is iteratively reconstructed and subtracted from the received signal. Parallel interference cancellation detector is compared in terms of performance with suboptimum linear detector, due to the reduced computational complexity.Other creators -
Languages
-
Spanish
-
-
English
-
-
Catalan
-
Recommendations received
2 people have recommended David
Join now to viewMore activity by David
-
🔐 Comment se passer de credentials statiques sur vos serveurs on-prem ? Le vrai problème : → Une clé JSON sur disque, c'est un secret statique…
🔐 Comment se passer de credentials statiques sur vos serveurs on-prem ? Le vrai problème : → Une clé JSON sur disque, c'est un secret statique…
Liked by David Sabater Dinter
Other similar profiles
Explore more posts
-
Imran Khan A.N
Fine-Tuning LLMs: Powerful, But Full of Hidden Pitfalls As more enterprises rush to operationalise GenAI, fine-tuning has quickly become the default hammer for every nail. But the truth is—fine-tuning is not a silver bullet. When done incorrectly, it introduces technical debt, operational fragility, and compliance risks that only surface once systems hit real production scale. At Cognizant’s AI Lab, we’ve spent the past year deeply researching and industrialising fine-tuning strategies—most recently exploring Evolution Strategies, read it here. https://lnkd.in/dNYvTbmH And through dozens of large-scale deployments, we’ve seen common pitfalls that teams often miss: ⚠️ Common Pitfalls in Fine-Tuning LLMs 1. Fine-tuning too early Teams jump straight into training models before exhausting prompt-engineering, RAG, and system-prompt optimisation. Result: unnecessary cost, slower iteration, and rigid models. 2. Using noisy or unstructured training data If your data has contradictions, tone inconsistency, or outdated content, the model will faithfully learn those flaws. LLMs amplify your data quality—not fix it. 3. Overfitting to niche internal language Fine-tuning on internal jargon without enough breadth can make the model brittle and hallucination-prone outside those patterns. 4. Forgetting evaluation frameworks Without gold-standard test sets and automated eval loops, you can never prove if the tuned model is truly better—or just different. 5. Lack of MLOps & observability Fine-tuned models must be monitored for drift, toxicity, hallucinations, and task-specific quality degradation. Production is where most fine-tuned models fail silently. 💡 What We’re Doing Differently At Cognizant, we’re at the forefront of deploying frontier models into production—securely, responsibly, and at enterprise scale. Our approach blends: 🔹 Multi-agent networks for reasoning, orchestration, and failover 🔹 Scalable architectures across Azure, AWS, and hybrid enterprise stacks 🔹 Fine-tuning frameworks that emphasise evaluation, safety, and continuous improvement 🔹 RAG + fine-tuning fusion patterns to deliver accuracy and adaptability 🔹 Compliance-ready pipelines for regulated industries Fine-tuning isn’t just about improving accuracy—it’s about building trustworthy, maintainable, and scalable AI systems that can survive in the real world. Jishnu Chatterji Tony L. Indranil Sen Deependra Mathur Ritesh Singh Babak Hodjat #AI #ArtificialIntelligence #GenAI #LLM #MachineLearning #EnterpriseAI #AgenticAI #MultiAgentSystems #AzureAI #MLOps #CognizantAI #AILab #FrontierModels #FineTuning #RAG #NeuroAI #AIAgents
1 Comment
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore More