Express YourSelf !

10 days Spoken Sanskrit workshops (Sanskrit Sambhaasan Shibir) held from July14 2025 to July 23 2025 at BEML Balaji Temple and at Skylark Arcadia Bangalore, by Samskrita Bharati Marathahalli bhaga, Bangalore, India. 

Spoken Sanskrit workshops - closing ceremony on 26th July 2025.

Sanskrit speech by chief guest Shri Manish Panchmatia

Greetings!

A warm welcome to everyone present—teachers, volunteers from Samskrita Bharati, students, and all those passionate about Sanskrit. After bowing down to each of you, I am honored to begin my speech.

First, let me ask: “How was the Spoken Sanskrit Workshop?” Was it “Good”? “Very good”? I’m certain it was a positive experience. Whenever Samskrita Bharati organizes a workshop, the teachers impart their knowledge with dedication, don’t they? Over the past ten days, each one of you demonstrated incredible determination to learn Sanskrit. No matter what, you attended all the classes and gave your best effort. Kudos to you all for your commitment and enthusiasm. You sang songs in Sanskrit, offered prayers, performed dramas, and narrated stories—all in Sanskrit. I am truly impressed by your efforts and progress.

As you may know, NASA called Sanskrit the most suitable language for computers. I won’t repeat those facts, but I do want to share a story about Sanskrit’s vast vocabulary. Just now, we sang about great poets—Vyasa, Bhasa, Kalidasa, Banabhatta—yet there is another renowned poet: Dandi, who composed “Dasakumaracharitam,” the story of ten princes. One prince, wounded in the lower lip, could not pronounce certain sounds: "pa," "fa," "ba," "bha," and "ma." Dandi skillfully chose synonymous words without those sounds for all of that prince’s dialogue. This was possible only because Sanskrit’s vocabulary is so rich.

Sanskrit grammar is exceptionally precise; in fact, the very meaning of the word “Sanskrit” is “well-formed” or “perfect.” Sanskrit has maintained its grammar rules since Vedic times—pronunciation, grammar, language rules—all unchanged. Even though new inventions like mobile phones have appeared, in Sanskrit, we create new words effortlessly. “Mobile phones” in Sanskrit is “Jangam Door-Vani”—a moving telephone. Artificial intelligence becomes “Krutrim Buddhi.” Thanks to Sanskrit’s structure, new words can always be formed. This habit of following rules brings discipline to our lives. 

We all are Indians. Sanskrit is deeply imprinted in our consciousness. Just as our childhood photos make us happy, speaking Sanskrit awakens ancient impressions within us—it fills the soul with joy. Throughout these ten days, you learned Sanskrit joyfully, supported by equally joyful teachers.

How do we gain these benefits? By giving your 100% effort to learning Sanskrit. What does 100% effort mean? Here is another story from the Mahabharata. Karna is renowned as the greatest giver. Once, Arjuna asked Krishna, "Why is Karna more famous for charity than my brother Yudhishthira?" Krishna replied, “Let’s see for ourselves.” Disguised as BRAHMINs, Arjuna and Krishna visited Yudhishthira and requested sandalwood for a yagna. Yudhishthir responded, "OK". Then, what did Yudhishthir do? He searched. Remember the story of the crow. The crow was searching for water. Here Yudhishthir is searching for sandalwood. He looked here . He looked there. Right side. Left side. Front side. Back side. He looked for Sandalwood everywhere. The sandalwood is nowhere. Then he went outside. Looked at the garden. He searched everywhere. The sandalwood is nowhere. He returned back and told BRAHMIN: "Sorry sir. The sandalwood is nowhere. I will surely make a little more effort to arrange for sandalwood. Please come tomorrow. I will give." BHRAMIN said, "OK".  Then, they visited Karna with the same request. Karna eagerly searched, and when he could not find sandalwood, he noticed his door was made of sandalwood. Without hesitation, he dismantled it and gave it to them. True charity is giving with 100% effort. It is the same with learning Sanskrit—when you give your wholehearted effort, you will gain both knowledge and discipline.

Further, speaking Sanskrit even helps with breathing exercises—“ma” and “ha” are frequently used, which naturally leads to “pranayama.”

This spoken Sanskrit workshop is just the beginning. Explore further—enroll in correspondence Sanskrit learning courses, Gita Sopanam etc. There are many Sanskrit books here, for sale. Buy them, read them and then, become Sanskrit teachers yourselves! Spread and promote Sanskrit to others just as your teachers did for you. You all know Swami Vivekananda. Right? "Yes". On 4th July, it was his death anniversary. On the last day of life, he taught Sanskrit to the students. You all know that? It is in his biography. Let us honor this legacy by teaching and learning Sanskrit.

Having spoken much about Sanskrit, let me turn briefly to our IT professionals. When we hear the word “language,” we often think of C, C++, Java, Python, NodeJS, ReactJS, and so on. Recently, I attended a "PyKrit" (Python + Sanskrit) workshop at Aksharam, Samskrita Bharati’s center.  This workshop was not for IT people. It was for Sanskrit scholars. I saw Sanskrit scholars name Python functions in Sanskrit, such as “YANA SANDHI” (a grammar topic). We truly can create software for Sanskrit grammar, using modern programming languages like Python, and even coding in Sanskrit.

Lastly, my wish: While we use AI/GenAI tools like ChatGPT, we mostly interact in English. Wouldn’t it be wonderful to have a large language model (LLM) in Sanskrit? Imagine asking questions in Sanskrit and receiving answers, back in Sanskrit, from AI tools, such as ChatGPT. That is my hope for the future.

Best wishes to everyone, and thank you all.

For AI application, we need automation of 

1. Data preparation
2. model tuning
3. Deployment
4. Maintenance and 
5. Monitoring

• Managing Dependency adds complexity. 

E2E workflow for LLM based application. 

MLOps framework

1. data ingestion

2. data validation

3. data transformation

4. model

5. model analysis

6. serving model

7. logging. 

LLM System Design

boarder design of E2E app including front end, back end, data engineering etc. 

Chain multiple LLMs together

* Grounding : provides additional information/fact with prompt to LLM. 

* Track History. how it works past. 

LLM App

User input->Preprocessing->grounding->prompt goes to LLM model->LLM Response->Grounding->Post processing + Responsible AI->Final output to user.

Model Customization

1. Data Prep

2. Model Tuning

3. Evaluate

It is iterative process

LLMOps Pipeline (Simplified)

1. Data Preparation and versioning (for training data)

2. Supervised tuning (pipeline) 

3. Artifact = config and workflow : are generated. 

- config = config for workflow

E.g. 

Which data set to use

- Workflow = steps 

4. Pipeline execution

5. deploy LLM 

6. Prompting and predictions

7. Responsible AI

Orchestration = 1 + 2 . Orchestration : What is first, then next step and further next step. sequence of step assurance. 

Automation = 4 + 5

Fine Tuning Data Model using Instructions (Hint)

1. rules

2. step by step

3. procedure

4. example

File formats

1. JSONL: JSON Line. Human readable. For small and medium size dataset. 

2. TFRecord 

3. Parquet for large and complex dataset. 

MLOps Workflow for LLM

1. Apache Airflow

2. KubeFlow

DSL = Domain Specific Language

Decorator 

@dls.component

@dls.pipeline

Next compiler will generate YAML file for pipeline

YAML file has

- components

- deploymentSpec

Pipeline can be run on

- K8s

- Vertex AI pipeline execute pipeline in serverless enviornment

PipelineJob takes inputs

1. Template Path: pipline.yaml

2. Display name

3. Parameters

4. Location: Data center

5. pipeline root: temp file location

Open Source Pipeline

https://us-kfp.pkg.dev/ml-pipeline/large-language-model-pipelines/tune-large-model/v2.0.0

Deployment

Batch and REST

1. Batch. E.g. customer review. Not real time. 

2. REST API e.g. chat. More like teal time library. 

* pprint is library to format 

LLM provides output and 'safetyAttributes'

- blocked

* We can find citation also from output of LLM

===========

vertexAI SDK

https://cloud.google.com/vertex-ai

BigQuery 

https://cloud.google.com/bigquery

sklearn

To decide data 80-20% for training and evaluation. 

Building AI/ML apps in Python with BigQuery DataFrames | Google Cloud Blog

===========

Examples: 

Istio, Kong, Envoy , Gluee , Trafeik, kgateway, Contour, NGINX, Kong Gateway and many more as per https://gateway-api.sigs.k8s.io/implementations/#gateway-controller-implementation-status

Protocols: gRPC, HTTP/2, and WebSockets

The structure of a Kubernetes Custom Resource Definition (CRD) or manifest file is referred to as an API. This is because it refers to the structure of the API in the Kubernetes control plane

Migration from ingress https://gateway-api.sigs.k8s.io/guides/migrating-from-ingress/#migrating-from-ingress

extension points

ingress has 2 extension points

1. annotations

2. resource end point

primary extension points in GW API:

1. External references

1.1 HTTP Route Filter

1.2 Backend Object Reference

1.3 Secret Object Reference

Here GW API reference 'external reference'

2. Custom implementations

e.g. The RegularExpression type of the 'HTTP Path Match'

3. Policies

A "Policy Attachment" is a specific type of metaresource

Here, Policy reference 'GW API'

• GW API is not API GW
• GAMMA (Gateway API for Mesh Management and Administration) initiative
• A 'waypoint proxy' is a proxy server deployed inside the mesh for E-W traffic. It can be (1) per namespace level destination services (2) few service(s) within a namespace as destination (3) destination services from multiple namespaces. 

- It is at cluster level. so no namespace

- Annotations at GatewayClassfor vendor specific

- It defines controller capabilities

2. Gateway

- Each Gateway defines one or more listeners, which are the ingress points to the cluster

- You can control which services can be connected to this listener (allowedRoutes) by way of their namespace — this defaults to the same namespace as the Gateway 

- Advanced featues like 

-- request mirroring, 

-- direct response injection, 

-- and fine-grained traffic metrics

-- Traffic spilt

- In Istio APIs, a Gateway configures an existing gateway Deployment/Service that has been deployed. In the Gateway APIs, the Gateway resource both configures and deploys a gateway

- one can attach HPA and PodDisuptionBudget to gateway deployment. 

3. HTTP Route: 

- any combinations of hostname, path, header values, HTTP method and query parameters.

• Paths (e.g., /headers, /status/*)
• Headers (e.g., User-Agent: Mobile)
• Query Parameters (e.g., ?version=beta)
• Methods (e.g., GET, POST)

- hostname (optional) at HTTP route shall match with hostname at Gateway->Listener->hostname

- A definition of the Gateway to use (in ParentRefs), is referenced by name and namespace

- The backendRefs that defines the service to route the request to for this match

- advanced pattern matching and filtering on arbitrary headers as well as paths.

1. RequestRedirect : E.g. Redirect HTTP traffic to HTTPS

2. URLRewrite

3. <Request|Response>HeaderModifier

4. RequestMirror

5. CORS

6. ExtensionRef for custom filter. E.g. DirectResponse

    filters:

    - type: ExtensionRef

      extensionRef:

       name: direct-response

       group: gateway.kgateway.dev

       kind: DirectResponse

- In the Istio VirtualService, all protocols are configured within a single resource. In the Gateway APIs, each protocol type has its own resource, such as HTTPRoute and TCPRoute.

- Traffic splitting is done by specifying multiple backendRef, with weight

- timeout, retry, sessionPersistence Session persistence, (= sticky sessions or strong session affinity), ensures that a client's requests are consistently routed to the same backend instance for the duration of a session. based on cookie or a header

- Route and Gateway can be in different namespace. If Gateway is defined with

    allowedRoutes:

      namespaces:

        from: Same

then Route and Gateway shall be in same namespace. We can have group of namespaces with label selector and specify those namespace using label at Gatway resource. 

    allowedRoutes:

      namespaces:

        from: Selector

        selector:

          matchLabels:

            self-serve-ingress: "true"

* 4. TLS Route

5. GRPCRoute

* 6. TCPRoute

* not v1, GA

Details: https://gateway-api.sigs.k8s.io/reference/spec/

If you are using a service mesh, it would be highly desirable to use the same API resources to configure both ingress traffic routing and internal traffic, similar to the way Istio uses VirtualService to configure route rules for both. Fortunately, the Kubernetes Gateway API is working to add this support. Although not as mature as the Gateway API for ingress traffic, an effort known as the Gateway API for Mesh Management and Administration (GAMMA) initiative is underway to make this a reality and Istio intends to make Gateway API the default API for all of its traffic management in the future.

https://gateway-api.sigs.k8s.io/mesh/

Gateway controller is for North South traffic. mesh controller is for East West traffic

7. ReferenceGrant: for cross-namespace reference. 

8. Inference Extension

K8s offers following mechanisms to optimize GPU usage, 

- time slicing, 

- Multi-Instance GPU (MIG) partitioning, 

- virtual GPUs, and 

- NVIDIA MPS 

for concurrent processing.

Effective GPU utilization means

- hardware allocation; 

- how inference requests are routed across model-serving instances. 

- how inference requests are load-balanced across model-serving instances. 

Simple load-balancing strategies often fall short in handling AI workloads effectively, leading to suboptimal GPU usage and increased latency.

Inference requests V/s traditional web traffic

- It often takes much longer to process, sometimes several seconds (or even minutes!) rather than milliseconds, 

- It has significantly larger payloads (ie, with RAG, multi-turn chats, etc). So a single request can consume an entire GPU, So making scheduling decisions far more impactful than those for standard API workloads. So, these requests need to queue up while others are being processed.

AI Models are stateful

- They maintain in-memory caches, such as KV storage for prompt tokens,

- They load fine tuned adapters like LoRA to customize response for specific user/organisation. 

So routing decisions are based on

- current state (in-memory caches, adapters)

- available memory, and 

- request queue depth.

So Inference aware routing through

8.1 Inference Model

- maps user facing model name to backend model

- traffic splitting between fine-tuned adapaters

- Priority based on real time interaction OR best-effort batch job

8.2 Inference Pool

- It is for platform operators managing model-serving infrastructure.

- a group of model-serving instances 

- specialized backend service for AI workloads.

- It manages

-- inference-aware endpoint selection, 

-- intelligent routing decisions based on real-time metrics such as 

--- request queue depth and 

--- GPU memory availability.

* Inference Pool is mapped with HTTP Route->backendRefs

* Inference Model has poolRef to link with Inference Pool

* Inference Pool has extensionRef (EPP = Endpoint picker) . If name for Inference Pool is xyz then extensionRef is "xyz-endpoint-picker" It is similar to K8s service, as it also has selector and target port

9. DirectResponse

10. Backend

For external endpoint

LTM Large Teleco Model : SoftBank is pioneer. Here is WhitePaper by GSMA https://www.gsma.com/get-involved/gsma-foundry/gsma_resources/white-paper-large-telecom-models/

Llama Nemotron Reasoning Model. Open source by NVIDIA on HF

https://www.nvidia.com/en-in/ai-data-science/foundation-models/nemotron/

https://arxiv.org/pdf/2505.00949

AI Factory is a specialized, integrated infrastructure designed to manage the entire AI lifecycle, from data ingestion to model training and deployment for real-time inference

AI Grid is a network of small, highly specialized AI communities. The members of AI Grid share their research work within these communities, initiate collaborations and establish fruitful connections for the future. https://lightning.ai/

https://ai-ran.org/

Building Blocks of the NVIDIA AI Aerial Platform: 

1. NVIDIA Aerial CUDA-Accelerated RAN

2. NVIDIA Aerial AI Radio Frameworks

3. NVIDIA Aerial Omniverse Digital Twin

Reference

NVIDIA GTC25: Telecom Special Address