As the founder of LLM Watch, my goal is to curate the most relevant Large Language Model (LLM) research and make it accessible to a broad audience. 

I distill complex technical topics into summaries that respect your time. 

Before embarking on this journey, I led Deep Learning R&D at one of Europe's largest telecom companies. There, I spearheaded numerous AI projects with a focus on Natural Language Processing (NLP) and Automatic Speech Recognition (ASR). 

With more than 6 years of experience in AI, I was there before the hype. 

I try to offer a healthy balance between shiny new things and what really matters.

Pascal Biese

Self-Evolving Memory:

A new framework from OPPO AI and LV-NUS lab now introduces memory systems that evolve their own architecture.

The paper draws a compelling analogy to human learning. Most AI agents today are like students who know how to take notes, but always use the same study method regardless of the subject. MemEvolve creates "adaptive learners" that dynamically adjust their learning strategy based on what they're trying to accomplish - memorizing facts for one task, abstracting patterns for another.

For this, they developed a dual-evolution process. The inner loop accumulates experience within a fixed memory architecture. The outer loop does something more interesting: it evaluates how well that architecture is working and proposes structural modifications. 

The system decomposes memory into four modules - Encode, Store, Retrieve, Manage - and uses performance feedback to diagnose bottlenecks and redesign components accordingly. 

The results? MemEvolve improves agent frameworks like SmolAgent and Flash-Searcher by up to 17% on challenging benchmarks. And the memory architectures evolved on one task transfer effectively to unseen benchmarks and different backbone models, suggesting the framework discovers generalizable principles rather than task-specific heuristics. 

In addition to the paper, they also released EvolveLab which unifies twelve existing memory systems under a common interface, making this a practical foundation for future research on self-improving agents. 
↓ 
𝐖𝐚𝐧𝐭 𝐭𝐨 𝐤𝐞𝐞𝐩 𝐮𝐩? Join my newsletter with 50k+ readers and be the first to learn about the latest AI research: llmwatch.com 💡

AI Lead at PwC <​/​> Daily AI highlights for 80k+ experts 📲🤗

MemEvolve: Meta-Evolution of Agent Memory Systems

A DeepSeek moment for AI memory? 

DeepSeek AI just published research suggesting that modern language models are burning valuable compute simulating memory lookups through expensive neural operations.

The problem is fundamental: language modeling involves two very different tasks. One is compositional reasoning - connecting ideas, logical inference. The other is knowledge retrieval - recognizing "Diana, Princess of Wales" or common phrases. Current Transformers have no native mechanism to simply look up static patterns, so they're forced to reconstruct them through multiple layers of attention and feed-forward networks.

The proposed solution is called Engram, a "conditional memory" module based on N-gram embeddings that retrieves static patterns in O(1) time. The key insight is that it's about finding the optimal balance between conditional computation (MoE experts) and conditional memory. The researchers discovered a U-shaped scaling law: pure MoE isn't optimal. Reallocating roughly 20-25% of sparse parameters to memory yields superior results. 

While memory was expected to help knowledge-intensive tasks (MMLU +3.4), the largest gains appeared in reasoning (BBH +5.0) and code/math (HumanEval +3.0, MATH +2.4). Mechanistic analysis reveals why: by offloading pattern reconstruction, Engram effectively deepens the network - shallow Engram layers produce representations equivalent to much deeper MoE layers. 

Perhaps most practically significant: the deterministic addressing allows offloading 100B-parameter tables to host memory with less than 3% overhead. 

This suggests that the next generation of sparse models might not just be about smarter routing - but about recognizing that some knowledge simply doesn't need to be computed at all.
↓ 
𝐖𝐚𝐧𝐭 𝐭𝐨 𝐤𝐞𝐞𝐩 𝐮𝐩? Join my newsletter with 50k+ readers and be the first to learn about the latest AI research: llmwatch.com 💡

A New Axis of Sparsity for Large Language Models

Meta's Super Intelligence Lab just published their solution to the training data problem. 

Training search-augmented LLMs typically requires extensive human-annotated questions and answers. As high-quality data becomes scarcer and more expensive, this dependency creates a significant bottleneck for developing capable AI research agents.

Dr. Zero introduces a clever proposer-solver feedback loop. Both models start from the same base LLM, but play different roles: the proposer generates increasingly difficult questions, while the solver learns to answer them using external search. 

The key insight is that as the solver improves, easy questions stop providing useful training signal, which forces the proposer to generate harder ones. This creates an automated curriculum that pushes both models forward without any human data.

The computational challenge with such approaches is typically prohibitive. Standard reinforcement learning methods require sampling many questions and many answers per question, creating nested costs that explode for multi-turn reasoning. Dr. Zero solves this through Hop-Grouped Relative Policy Optimization (HRPO), which clusters questions by their structural complexity (number of reasoning "hops" required) and computes advantages within these groups. This reduces sampling overhead to roughly one-fourth of standard methods while maintaining performance. 

The data-free Dr. Zero matches or exceeds fully supervised search agents across seven QA benchmarks, with up to 14.1% improvements on complex multi-hop reasoning tasks. 

This suggests that complex reasoning and search capabilities can emerge purely through self-evolution. 
↓ 
𝐖𝐚𝐧𝐭 𝐭𝐨 𝐤𝐞𝐞𝐩 𝐮𝐩? Join my newsletter with 50k+ readers and be the first to learn about the latest AI research: llmwatch.com 💡

Self-Evolving Search Agents without Training Data

LLM-as-a-Judge -> Agent-as-a-Judge

Using LLMs to evaluate other LLMs seemed like a clever shortcut. But single-pass evaluation is hitting its limits: we need evaluators that can actually verify claims, not just assess how plausible an answer sounds. 

Traditional LLM judges are passive observers. They grade based on linguistic patterns without checking if the code runs, the math is correct, or the facts hold up. 

A new comprehensive survey traces the shift to Agent-as-a-Judge, where evaluation becomes an active, multi-step process. Instead of one model making a snap judgment, agentic judges decompose evaluation into subtasks, invoke external tools like code interpreters and search engines to verify claims, and use multi-agent debate to neutralize individual model biases. 

The researchers identify three developmental stages: procedural agents following fixed workflows, reactive agents adapting based on intermediate feedback, and self-evolving agents that can refine their own evaluation criteria on the fly. Applications already span code verification, fact-checking, medical assessment, legal judgment prediction, and financial research analysis. 

When AI systems are used to generate training signals for other AI systems, the precision of the judge determines everything downstream. Can we build evaluators that know when they need to verify rather than guess?
↓ 
𝐖𝐚𝐧𝐭 𝐭𝐨 𝐤𝐞𝐞𝐩 𝐮𝐩? Join my newsletter with 50k+ readers and be the first to learn about the latest AI research: llmwatch.com

A Survey on Agent-as-a-Judge

What if you could surgically edit an LLM's knowledge by swapping a single vector?

A new architecture from CMU and Meta AI makes this possible, while also solving some of the thorniest problems in scaling transformers.

Scaling language models has traditionally meant more compute. Mixture-of-Experts (MoE) promised a workaround by activating only a subset of parameters per token, but came with notorious training instability, load balancing headaches, and communication overhead that grows with expert count.

Instead of dynamic routing to experts, STEM (Scaling Transformers with Embedding Modules) replaces the FFN up-projection with a static, token-indexed embedding lookup. Each token has its own embedding per layer, while the gate and down projections remain shared. No routing decisions, no load balancing losses, no cross-device communication for expert selection. 

The results? 3-4% accuracy gains over dense baselines, with particularly strong improvements on knowledge-intensive benchmarks (9-10% on ARC-Challenge and OpenBookQA). Training remains stable where MoE would show loss spikes, and the architecture naturally activates more parameters as context length grows. 

Because each embedding maps to a specific token, you can literally swap "Spain" for "Germany" at the embedding level and watch the model output "Berlin" instead of "Madrid", without changing the input text at all. 

If interpretability and controllability matter as much as raw performance, STEM suggests we may have been overcomplicating sparse scaling all along.
↓
𝐖𝐚𝐧𝐭 𝐭𝐨 𝐤𝐞𝐞𝐩 𝐮𝐩? Join my newsletter with 50k+ readers and be the first to learn about the latest AI research: llmwatch.com

STEM: Scaling Transformers with Embedding Modules

The next big step in LLMs?

A new inference technique from Massachusetts Institute of Technology called Recursive Language Models (RLMs) is rethinking context windows entirely.

The core problem is well-known: even frontier models like GPT-5 suffer from "context rot" - performance degrades quickly as prompts get longer, regardless of the technical context limit. Summarization helps but loses critical details. Retrieval misses complex reasoning patterns. Have we been trying to make models see more, when perhaps the answer is to make them see differently?

RLMs treats the prompt not as direct neural network input, but as an external object the model interacts with programmatically. The prompt is loaded as a variable in a Python REPL environment, and the LLM writes code to peek into it, decompose it, and recursively call itself over smaller snippets. Same interface as a regular LLM, radically different execution.

On information-dense tasks where GPT-5 scores below 0.1%, RLMs achieve 58%. On multi-hop research questions spanning 6-11M tokens, RLMs hit 91% accuracy while costing less than feeding the full context would. Crucially, performance degrades far more gracefully as complexity scales - the approach handles inputs two orders of magnitude beyond native context windows.

This suggests that scaling context is not just an architecture problem but also an inference problem. RLMs demonstrate that letting models reason about their input symbolically rather than processing it neurally could be a promising new direction.

If this approach generalizes, we may be looking at a new axis for scaling language model capabilities entirely.
↓ 
𝐖𝐚𝐧𝐭 𝐭𝐨 𝐤𝐞𝐞𝐩 𝐮𝐩? Join my newsletter with 50k+ readers and be the first to learn about the latest AI research: llmwatch.com 💡

Recursive Language Models

The latest reasoning models fail on puzzles with just 1 unknown cell.

Yet they solve complex ones with dozens of steps.

New research reveals why: these models might be pattern matching, not actually reasoning.

Hierarchical Reasoning Models (HRMs) crush benchmarks, outperforming LLM-based reasoners by wide margins. But researchers dug into the mechanics and found something troubling.

Think of it like a student who memorizes exam answers. They ace questions they've seen before but stumble on trivially easy variations.

The researchers identified three failure modes:

- HRMs fail on extremely simple puzzles (single unknown cell) when patterns violate their training distribution
- Models lack the "fixed point property" - they can't recognize when a problem is already solved
- Performance doesn't correlate with puzzle difficulty the way true reasoning would

This matters because we're deploying these models assuming they generalize.

If your reasoning model solves a 50-step problem but fails a 2-step variant, it's not reasoning - it's sophisticated pattern matching.

Before trusting HRMs in production, test them on deliberately simple edge cases. The failures will tell you more than the successes.
↓
𝐖𝐚𝐧𝐭 𝐭𝐨 𝐤𝐞𝐞𝐩 𝐮𝐩? Join my newsletter with 50k+ readers and be the first to learn about the latest AI research: llmwatch.com 💡

Are Your Reasoning Models Reasoning or Guessing?

Kloudstac

Kloudstac is a AGI (Artificial General Intelligence) Acceleration Platform startup born in the ChatGPT era, helping Fortune 2000 enterprises in improving adoption & value realization from AGI via our learn & build playgrounds and use case pilot accelerators.

We are the World's leading provider of high quality AI talent transformation solutions (as part of 1st leg of our AGI Acceleration agenda) incl largest portfolio of AI upskilling programs (30+) in the Industry and 500+ labs/ playgrounds across personas in Generative and Agentic AI. 

Our unique approach offers Enterprises a Holistic, Continuous, Accelerated & scalable AI Talent Transformation solution that is unmatched in the Industry.

Delivered over 150 Generative and Agentic AI programs in last 2.5 yrs. We partner with Fortune 2000 Enterprises and top 30 leading Edtech/ Corporate and B2C education providers globally (across 5 countries) for servicing their customers on AGI/ Agentic & Generative AI talent development. 

Our offerings:
1. AGI Playground platform with learn, build labs & personalized AI tutor supporting each learner in their AGI skill development journey, coupled with our Agentic AI Copilots/ Pilot Accelerators
2. Agentic AI Industry Solutions and AGI Strategy & Value discovery (use cases & pilots, implementations)- our consulting services
3. 30+ AGI / Agentic AI training programs for B2B, B2C - across technologies, Industries, Business functions

Website: https://www.kloudstac.com

My specialization:
1. Mentor/ Advisor (AI/AGI) for Business & Technology Leaders across Industries, GenAI/ Agentic AI Architect programs across AWS, GCP, Azure, OCI (delivered ~100 batches till Jun 2025)
2. AI Use cases incl AI Agents across Industries, horizontals incl frameworks, use case discovery, detailing
3. Architecture & development- RAG, AI Agents, Finetuning: OpenAI, Anthropic, Google, Microsoft, NVIDIA, Mistal, Meta stack
4. LLMOps & MLOps- GCP Vertex AI, AWS Bedrock & Sagemaker, Azure AI Studio

CEO & Product GTM Leader, Generative/Agentic AI/AGI Advisor & Skills Trainer

Globussoft Technology

Building Generative AI (GPT2/3.5) driven content generation app.
Responsible for business-as-usual/ ops, re-architecting/ modernizing, new features, improvements, growing revenue on 3 SAAS products in digital marketing, adtech, workforce productivity & engagement (empmonitor.com, poweradspy.com, socinator.com).
Tech stack: GCP- MongoDB, Firestore, cloud functions, GCS, Bigquery, GKE-Kubernetes, Compute- VMs, containers, docker, Cloudsql-RDS, Elasticsearch, APIGEE; AWS- S3, EC2, Lambda, RDS-mysql, documentdb, API Gateway, athena; NextJS, NodeJS, Polygon L2 scaling with solidity-ethereum
Achievements:
1. Restructured product organization, cutting 30% cost, optimized engg team. This includes reducing cloud costs by 70-80% in 8 months. 
2. As architect, re-engineered/ created to-be design for EMPMonitor.com on AWS, later on GCP and poweradspy.com on GCP. Used best-in-class tech stack - serverless and containerized microservices.
3. Changed/ modernized the obsolete/in-consistent product strategies for EMPMonitor.com (employee monitoring to Workforce productivity & engagement platform), poweradspy (updated to- "Discover social media ads + content strategy of competitors") with 18 months product roadmap in line with market reqs/in-demand use cases.
4. Improved processes & practices/mix across sales, marketing, ops leveraging past experience in startups, large enterprises. Cut down marketing costs by half within 3 months.
5. Executed ISV partnership with Google cloud platform giving global access to customers and GCP sales network via GCP Marketplace
6. Defined, optimized product roadmaps, pricing & GTM strategy on continuous basis and provided needed leadership & mentorship to various teams.

Chief Product & Technology Officer

Skills: Product Design - UI/UX, business workflows, use case development
Product Engg: AWS, GCP - OLTP, OLAP backends; AI - NLP, CV, Machine teaching; Data engg: Snowflake, Databricks; Web3.0/ Blockchain: Polygon, Solana

Helping tech startups on their product journeys:

As Product advisor:
1. Business model & product design- Edtech startup focused on rightskilling engg students
2. LiveCommerce startup on product platform & tech design
3. Design- Fintech BNPL (Buy now pay later) platform

Tech/ Engg Advisor:
1. HR Tech co - App platform (Azure) architecture & design assessment, improvement, new capabilities
2. Edtech platform- MVP development
3. App platform (AWS) architecture & design assessment for a Travel tech co
4. Solutioning for "machine operator automation" use case using Deep RL, Machine teaching for manufacturing.

Startup Advisor- Product & Technology

GoBaskt Inc

GoBaskt (US headquartered, http://gobaskt.com) is a tech product startup offering a hyperlocal commerce platform for local businesses, retailers & CPG Brands. We are helping businesses acquire & retain customers and improve revenue within local area of 5-6 miles with our unique technology.
For local businesses, we bring modern technology, that only large brands can afford, on a value based pricing model.

For consumers, we offer an intelligent virtual shopping assistant app that uses AI, multi channel engagement, analytics in new ways, in order to create unique interaction experiences between individuals and businesses for commerce, marketing and service.  It helps consumers to save time, money & efforts, while shopping for services & products.

Achievements:
1. Developed product vision-mission, value proposition, use cases, UI-UX/ business workflows, pivoted to more valuable use cases during covid times and built them fast.

2. Built technology strategy, designed, developed, deployed complex full stack web and mobile apps  with NextJS-React for web apps (migrated from angular), backend with python-nodejs (migrated from Java) + AWS Lambda, Containerized microservices with AWS Fargate/ ECS and EKS/ Kubernetes, Snowflake, S3, Athena, Kinesis/ Kafka, AWS Step functions, mobile apps with IOS native, Android native (Kotlin), databases- dynamodb, mongodb, postgresql, DevOps- Github-AWS Codebuild, codepipeline, codedeploy. Multiple use case built and deployed in production over 2 yrs.

3. Ran engineering/ operations frugally with a low budget esp during COVID time when primary customer market got disrupted. Bought revenue/ sales.

Co-Founder, Chief Product & Technology Officer

TheCyborgs

1. Academic level program to enable emerging tech skill development via high quality hands-on projects and mentorship from experienced professionals. So far, 30+ students across few engineering institutions benefited from the program.

2. Global virtual "Artificial Intelligence" Professional Community comprising 150+ AI practitioners, with the aim of mastering concepts, industry use cases, code implementation and practical hands-on innovation project based application of emerging AI techniques. 

Focus areas- Computer Vision, NLP, Predictive/ Prescriptive/ Descriptive analytics, AWS & Azure, Chat & voice bots, Web & mobile apps, Augmented Reality.

Sample recent topics (AI):
1. Transfer learning for NLP with Google BERT
2. LSTMs, GRUs, Attention models & transformer architecture
3. Capsule nets
4. GAN Implementations
5. Yolo v3 object detection
6. Deepspeech architecture, CTC, Attention models for ASR
7. Siamese Networks
8. Brain Computer Interfaces
9. Deep Reinforcement learning with OpenAI, Tensorflow
10. Recommendation models

Founder (Artificial Intelligence Tech Research & Innovation Community- Industry-Academia)

Kore.ai

25+ engagements (POCs/ pilots/ implementations/ running large programs) (CPG, Telcos, Banks) as program/project mgr, BA, Solution architect (includes pre-sales work). Clients- Loreal, Verizon, Citibank, Deutsche Bank, Philips Lighting, BB&T Bank, Nestle, P&G, Yahoo Sports, ABInBev, Colgate, etc.

Built Professional Services/ Delivery organization for North America- Sept 2017 to Jun 2018. 

Managed solution design, development & delivery in multiple strategic customers- bot solutions roadmap, use cases discovery/ new projects/ generate additional business, help improve operational processes & capabilities of deployed bot solutions.
Hands-on bot building on Kore, facebook, amazon alexa, google assistant, IOS & Android apps.
Hands-on/ deep expertise on AI engines in production: NLP- Supervised & unsupervised ML, Computational Linguistics/ Fundamental meaning, Knowledge graph.
Expertise in discovering/ proactively developing 20+ net new use cases.

As part of Product Mgt (Monetization, New Products), my responsibilities:

1. Executing services (delivery, solution design, dev, deploy)- POV Pilots, Implementations and improving customer satisfaction & value creation from using kore products
2. supporting license sales cycles & proposals as functional and product SME/ pre-sales architect
3. working with product management on new products (idea-to-MVP), product improvements/ bug fixes/ enhancements/ new functionalities (includes providing detailed requirement, conceptualizing user stories).
4. content creation- blogs, articles, case studies, etc
5. contributing to SI and OEM alliance development as SME
6. Conceptualized, prototyped new product ideas.
7. demos- industry use cases, developing day-in-the-life scenarios, conversation flows.
8. setting up/ optimizing chatbot programs (covering entire bot lifecycle) in large enterprises
9. Work with product engg daily/weekly on technical issues/ bugs, limitations/ new technical features.

Director- Outbound Product Management

Capgemini

Key focus areas (Digital Transformation initiatives- practice sales, advisory, delivery, IP dev): 
1. AI/ Cognitive & Intelligent Automation- chat & voice bots, conversational analytics & NLG, deep learning, computer vision, NLP based automation (OCR, intent & entity extraction from email text)
2. Big data analytics- SPARK, R, python, HADOOP HDFS
3. Industrial & Consumer IOT (Device & Sensor data mgt, IOT protocols, Edge & Cloud analytics, algorithms, real time operational dashboards, KPIs/ metrics, asset maintenance& monitoring )
4. Blockchain technology driven Industry solutions for product supply chain based sectors- CPG, Retail, Pharma, O&G
5. CXO Cockpit/ Digital Boardroom analytics - 360 view of corporate performance & key driver analysis

Partial hands-on exposure to 
a) Big data technologies:  SPARK, NoSQL DB - MongoDB
b) Statistical and Machine learning tools/ languages with R, Python (Scikit learn, numpy, pandas)
c) Blockchain tools @ Ethereum
d) Conversational apps with Enterprise AI Platforms- chatbots/ virtual intelligent assistants
e) Formerly- SAP HANA, SAP BPC, SAP BI, Analysis Office & BOBJ Cloud
f) Cognitive/ ML APIs- Google cloud platform, Microsoft Azure, IBM Watson
g)  Deep Learning with Google Tensorflow

SME on emerging solution spaces in Digital Marketing & Marketing driven commerce- Account Based Marketing, Personalized & contextual multi-channel multi-wave campaigns, dynamic customer profile, segmentation, predictive analytics, social listening, location based marketing, mobile loyalty programs, use of AI/Cognitive technologies for Digital Transformation across front office & back office business processes and business model re-configurations in select industries.

Hands-on exposure on AI frontiers- Meta learning, Siamese networks, Generative adversial networks. GAN, GRUs & LSTM, etc.

Senior Manager, Emerging Digital Technologies (Insights & Data)

Developed global SAP HANA Practice with GTM collaterals/assets, POVs, service offerings, case studies, solution accelerators, relationships with SAP Labs & SAP/ Hybris Field Sales.

Managed HANA requirements (solutioning, proposals) on BI-Native HANA, BW-HANA, BPC-HANA, S/4HANA, Fiori/UX, HANA Cloud, Hybris Marketing across 20+ customers & prospects.

Drove demand generation programs on HANA (new pursuits & account mining initiatives) with sector sales leaders.

Key contributions/achievements:
1. Achieved 500% growth in HANA SMEs headcount in 1.2 yrs.
2. Built 23+ solution prototypes/ demos on HANA across applications
3. Achieved approx. 800% growth in HANA Case studies in 1.2 yrs.
4. Achieved 270% growth in revenue from HANA Services in 1.2 yrs. 
5. Delivered multiple HANA Engagements- Implementation, Pilot, Advisory successfully.
6. Developed 30+ HANA Collaterals- capability decks, reference architectures, POVs, solution models, etc
7. Built multiple new HANA Cloud alliances as part of building HANA Cloud practice and developed highly differentiated joint service offerings.

SAP HANA Practice & GTM Head, North America

ITC Infotech USA Inc

Worked with Sales Teams, Partners and customers/ prospects on project support/delivery oversight, proposal/ requirement solutioning in Business Analytics & EPM. Applications include - BPC, BW, HANA- Analytics. Identification & development of various sales channels, drive capability development initiatives.

Built competencies on Consolidation/ Group Close based on IFRS Standards (IAS 27, 28, IFRS 3,10,11)

Managed lead generation & deal pipeline in North America on SAP BPC, BW, BusinessObjects, HANA by working with Geo Sales, Partners and Inside Sales.

Oversee delivery/ solutioning for offshore projects on BPC/ BPC-BW for North American Customers.

Developed well rounded exposure in BPC - Planning : Sales, Material, Production/ S&OP, Direct & Indirect Costs including HR Headcount & payroll cost, CAPEX, Balance Sheet, Income Stmt, Cash Flow Stmt, Machine level Capacity Planning, Plan data retraction to ECC, FICO Processes. Discrete & Process Manufacturing.

Bought improvements in Sales Process for SAP Business Analytics & EPM in following areas:
1. Prospecting, Lead Generation, Sales Campaigns
2. Initial Engagement and Positioning Services/ Value Offerings

Driving Practice building & Capability Dev. in BusinessObjects Reporting and Advanced Analytics (Descriptive/ Predictive/ Prescriptive on Big Data) on SAP Predictive Analysis/ Infinite Insight, HANA, R-HANA-HADOOP Integration, SAP's Advanced Analytics RDSes for Sales, Marketing, Manufacturing, Retail/ CPG

PreSales & Solution Delivery Head - SAP Business Analytics & HANA, North America

- Built BPC and HANA Practice since inception at ITC Infotech. Developed SAP BPC and HANA In-Memory Analytics Capabilities, Offerings, New Solutions, GTM. Work on PreSales, Implementations, Consulting Engagements including True Profitability/ Costing, Shared Services Costing, Driver based Financial Planning
- PreSales on SCPM-HANA, Demand Signal Management on HANA
- Evolved Hub-n-spoke model of HANA Practice, did internal selling and put into practice
- Expanded sales channels and utilized them for lead generation

Capabilities Developed
1. BW and BPC Database Migration to HANA
2. HANA COPA Accelerator
3. HANA Demand Signal Management
4. HANA Information Models/ Custom Data-mart and consumption into BW data models
5. HANA Value Assessment and Proof of Value Consulting Offering

Demo Solutions designed and executed/ in-progress on HANA:
1. Simple and Complex Reporting Scenarios with high data volume
2. Mixed Scenarios – Consuming HANA Information models in BW and vice versa
3. Real Time Reports and Dashboards using SLT Replication in HANA Information Models
4. Data Extraction into HANA using DXC (Direct Extractor Connection)
5. Inventory Reports & Dashboards on BOBJ-HANA with 3 million records, performance comparison with similar scenario on BW – Traditional Database
6. Market Research and POS data Analytics in Demand Signal Management-HANA

Practice & PreSales Head - SAP BPC, HANA Solutions

Philips Electronics Limited India

Working as a SME in Supply Chain Planning/APO and Finance ( Profitability and Cost Analytics) for SAP BW. Was part of BW projects on Supply Chain Planning, Profitability & Cost Mgmt, took several handovers of several projects during transition to Support, Managed Team for delivery on Incidents and Data Load Monitoring, Involve in Technical Incident Resolution and Data Load Monitoring, Knowledge Management and Training Program Dev. for SAP BW Team.

Developed Functional knowledge on supply chain planning and dimensional profitability & cost allocation mechanisms.

Sr Technical Lead/ SME for Supply Chain Reporting & Analysis

Accenture

SAP Business Intelligence Team Lead, & Techno-Functional Consultant, Functional SAP FICO. MM Consultant, BusinessObjects Reporting & Analysis

SAP Business Intelligence Consultant and Team Lead

21+ yrs of experience in B2B and B2C Software Product Lifecycle (SAAS, On-premise/ run anywhere)- Ideation, MVP, Post revenue/ growth stage, late stage, rearchitecting/ modernizing, high scale (retail, telco) & multi-country complex production environments, customer implementations, customer onboarding, presales/ sales support, product roadmap & release mgt, multi-product operations, product design (UI-UX), product analytics incl customer journey mapping.

10+ yrs in AI/ML incl 4 yrs in Generative AI, 2 yrs in Agentic AI & RAG. 35+ AI Projects delivered as Architect incl 15+ production deployment, part of engineering/ product mgt in 4 AI powered products.

Worked on products with peak load stats: MAU of upto 3 mil users, 40-50k DAUs, 1000-1500 transactions per sec (tps).
Max team sizes handled - 300

Core Experience :

1. Agentic & Generative AI/ Artificial General Intelligence (3 yrs)- LLMs, Vector databases, Advanced design patterns & strategies- AI Agent development, AWS Bedrock, GCP Vertex AI, Azure AI Studio, Advanced RAG techniques, evaluation & accuracy improvement, finetuning- Supervised, RL.

2. Classic ML- Supervised, unsupervised learning
3. Data Engg & Analytics- Google Bigquery, Dataflow, Cloud composer (Airflow), Spark, Kafka
4. Cloud Infra & DevOps (AWS, GCP, Azure)

Domains- 
Supply Chain: S&OP/ IBP, New product introduction, demand sensing & shaping, O2C, supply chain control tower, SCOR, MRP, Master prod scheduling & RCCP, P2P, MD, asset ops.
Customer experience (Marketing, Sales, Service): multi-channel commerce, Customer 360, journey mapping, campaign, loyalty, planning, analytics (attribution, ROI, churn, market basket analysis), recommendations, lifetime value

FP&A: Planning/ Budgeting/ Forecasting, analytics (profitability, balance sheet, cash flow, EVA/ Economic profit, value creation), subsidiary & group close/ consolidation, activity based costing, MD.

Industries: Retail/Ecom, CPG, O&G, BFSI

8 yrs Entrepreneurial Experience: co-founder/ partner in 5 startups (past, current).
Product & CTO advisor to multiple tech startups.

taplio

The Battle

Pascal Biese

Abhiroop Roy

Followers

Posts

Engagements

Likes

Performance ScorePascal Biese's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Pascal Biese

Abhiroop Roy

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Pascal Biese's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScorePascal Biese's LinkedIn Performance