Services
Courses by Kumaran
-
Business Analytics Foundations: Descriptive, Exploratory, and Explanatory Analytics48m
Business Analytics Foundations: Descriptive, Exploratory, and Explanatory Analytics
By: Kumaran Ponnambalam
-
Agentic AI for Developers: Concepts and Application for Enterprises1h 31m
Agentic AI for Developers: Concepts and Application for Enterprises
By: Kumaran Ponnambalam
Articles by Kumaran
-
Are You a Java Developer ? Start thinking Scala
Are You a Java Developer ? Start thinking Scala
Java, the legendary programming language came into existence about 20 years ago. It had solutions to a number of…
-
Next Gen Contact Centers: Tech. Enabled, Context Driven, Customer Focused
Next Gen Contact Centers: Tech. Enabled, Context Driven, Customer Focused
The traditional idea of a contact center – a large room; cubicles arranged in orderly fashion; agents sitting before a…
1 Comment -
Predictive Analytics for Unified Communications
Predictive Analytics for Unified Communications
More and more organizations today are moving to unified communications (UC) platforms for better communications within…
Activity
22K followers
-
Kumaran Ponnambalam shared this𝗜𝗳 𝘆𝗼𝘂𝗿 𝗔𝗜 𝗮𝗴𝗲𝗻𝘁 𝗻𝗲𝗲𝗱𝘀 𝗛𝘂𝗺𝗮𝗻-𝗶𝗻-𝘁𝗵𝗲-𝗟𝗼𝗼𝗽, 𝗱𝗼 𝘆𝗼𝘂 𝗸𝗻𝗼𝘄 𝘄𝗵𝗲𝘁𝗵𝗲𝗿 𝗛𝗜𝗧𝗟 𝗶𝘀 𝗰𝗿𝗲𝗮𝘁𝗶𝗻𝗴 𝘃𝗮𝗹𝘂𝗲 𝗼𝗿 𝗷𝘂𝘀𝘁 𝗮𝗱𝗱𝗶𝗻𝗴 𝗱𝗲𝗹𝗮𝘆 𝗮𝗻𝗱 𝗵𝗶𝗱𝗱𝗲𝗻 𝗵𝗲𝗮𝗱𝗰𝗼𝘂𝗻𝘁? Different Gen AI agents and applications need different levels of HITL. Adding HITL into the workflow ensures 𝘀𝗮𝗳𝗲𝘁𝘆 of outcomes whether its about accuracy, policy or hallucinations. But it adds 𝗱𝗲𝗹𝗮𝘆𝘀 when awaiting human feedback. Time spent in human review also adds 𝗰𝗼𝘀𝘁𝘀. Organizations need to balance between these conflicting elements based on their use case. This means, we need to measure the true value added by HITL and then strive to improve it for effective business outcomes. In this article I break HITL down as an ROI problem and propose a simple metric stack: 1. Net Value (risk prevented minus review + delay cost) 2. Value-per-Cost (efficiency) 3. An F1-style Balanced Score (effectiveness penalized by cost/latency) Read more about it here: https://lnkd.in/gzRErc-R
-
Kumaran Ponnambalam shared thisExcited to speak on this webinar and share practical, enterprise-grade patterns for proving AI ROI and controlling risks in production . Registration link: https://lnkd.in/giK43RdBKumaran Ponnambalam shared thisThe demos went well. Six months later, leadership wants to know what it returned. Most AI teams are not ready for that question. Not because the work was not done. Because the accountability structures were never built alongside it. That is not a technology problem. That is a structural one. And it is the conversation most organisations are quietly avoiding. On 10 June, we are having it out loud. Three practitioners who have built these systems in the real world, sharing what actually works. Sowmya Vivek, AI Solutioning Consultant and GenAI Strategist, Merit Data and Technology, Kumaran Ponnambalam, Principal ML Engineer, Cisco & Tharun Mathew, Head of Data and AI Solutions, Merit Data & Technology. 30 minutes. No pitch. No theory. Just the blueprint for making AI investment answerable. 10 June 2026 | 4:00 PM BST | 8:00 AM PDT | Live + Q&A Can't make it live? Register anyway. We will share the recording. Link to register in comments. #AI #EnterpriseAI #AIStrategy #DigitalTransformation #Webinar
-
Kumaran Ponnambalam shared this𝗪𝗵𝗮𝘁 𝗵𝗮𝗽𝗽𝗲𝗻𝘀 𝘄𝗵𝗲𝗻 "𝗮𝗻𝘆𝗼𝗻𝗲 𝗰𝗮𝗻 𝗯𝘂𝗶𝗹𝗱 𝗮𝗻 𝗮𝗴𝗲𝗻𝘁" 𝗯𝗲𝗰𝗼𝗺𝗲𝘀 “𝘄𝗲 𝗵𝗮𝘃𝗲 𝟮𝟬𝟬 𝗮𝗴𝗲𝗻𝘁𝘀 𝗮𝗻𝗱 𝗻𝗼 𝗼𝗻𝗲 𝗸𝗻𝗼𝘄𝘀 𝘄𝗵𝗶𝗰𝗵 𝗼𝗻𝗲𝘀 𝗮𝗿𝗲 𝘀𝗮𝗳𝗲”? A recent arXiv paper proposes a direct answer: treat agent governance like a maturity journey, not a one-off checklist—because unmanaged agent sprawl becomes a business operations problem. https://lnkd.in/gUc5QPth. The paper argues enterprises are already seeing agent sprawl patterns like functional duplication, shadow agents, orphaned agents, permission creep, and unmonitored delegation chains - each with measurable cost/risk impact. It recommends • A 𝗺𝗮𝘁𝘂𝗿𝗶𝘁𝘆 𝗹𝗮𝗱𝗱𝗲𝗿 (L1 ad-hoc to L5 optimized) with concrete controls like registry, IAM, observability, policy enforcement, human oversight, and governance-as-code. • A 𝘀𝗽𝗿𝗮𝘄𝗹 𝘁𝗮𝘅𝗼𝗻𝗼𝗺𝘆 + 𝗰𝗼𝘀𝘁 𝗺𝗼𝗱𝗲𝗹 to quantify redundancy, security/compliance exposure, and operational overhead. • A 𝗴𝗼𝘃𝗲𝗿𝗻𝗮𝗻𝗰𝗲 𝗱𝗼𝗺𝗮𝗶𝗻 𝗰𝗮𝘁𝗮𝗹𝗼𝗴 that covers lifecycle, runtime governance, portfolio governance, and continuous improvement. In their simulation-based evaluation, higher maturity levels show large reductions in sprawl and risk incidents while improving effective task completion.
-
Kumaran Ponnambalam posted this𝗪𝗵𝗲𝗻 𝗽𝗲𝗼𝗽𝗹𝗲 𝘀𝗮𝘆 "𝗼𝗽𝘁𝗶𝗺𝗶𝘇𝗲 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗔�� 𝗰𝗼𝘀𝘁 𝗮𝗻𝗱 𝗹𝗮𝘁𝗲𝗻𝗰𝘆", 𝘄𝗵𝗮𝘁 𝗹𝗮𝘆𝗲𝗿 𝗮𝗿𝗲 𝘁𝗵𝗲𝘆 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝘁𝗮𝗹𝗸𝗶𝗻𝗴 𝗮𝗯𝗼𝘂𝘁? Most teams focus on the model or GPU utilization, but the real wins come from treating the stack as a set of levers across layers: 𝗮𝗴𝗲𝗻𝘁 -> 𝗰𝗼𝗻𝘁𝗲𝘅𝘁 -> 𝗺𝗼𝗱𝗲𝗹 -> 𝗶𝗻𝗳𝗲𝗿𝗲𝗻𝗰𝗲 𝘀𝗲𝗿𝘃𝗲𝗿 -> 𝗿𝗼𝘂𝘁𝗶𝗻𝗴 -> 𝗚𝗣𝗨 𝗰𝗹𝘂𝘀𝘁𝗲𝗿. Here’s a simple way to think about optimization by layer: 1. 𝗔𝗴𝗲𝗻𝘁: improve planning, tool-use, verification, and policies; cut cost by reducing steps/retries/loops; cut latency by parallelizing and early-stopping. 2. 𝗖𝗼𝗻𝘁𝗲𝘅𝘁/𝗥𝗔𝗚: improve retrieval and grounding; cut cost by reducing tokens via filtering/compression; cut latency with shorter prompts and caching. 3. 𝗠𝗼𝗱𝗲𝗹: use stronger or domain models where needed; cut cost by using smaller models for simpler tasks; cut latency by using faster models for simpler tasks. 4. 𝗜𝗻𝗳𝗲𝗿𝗲𝗻𝗰𝗲 𝘀𝗲𝗿𝘃𝗲𝗿: protect quality where it matters; cut cost via batching/concurrency/quantization; cut latency via lower queue time, TTFT optimization, and streaming. 5. 𝗥𝗼𝘂𝘁𝗶𝗻𝗴: choose the right model/pipeline per request; cut cost with cache-aware routing and bin packing; cut latency with priority queues and traffic pools. 6. 𝗚𝗣𝗨 𝗰𝗹𝘂𝘀𝘁𝗲𝗿: enable bigger models safely; cut cost with utilization and autoscaling; cut latency with headroom and isolation. 𝗧𝗵𝗲 𝗯𝗲𝘀𝘁 𝘁𝗲𝗮𝗺𝘀 𝗱𝗼𝗻’𝘁 𝗷𝘂𝘀𝘁 𝘁𝘂𝗻𝗲 𝗮 𝗺𝗼𝗱𝗲𝗹, 𝘁𝗵𝗲𝘆 𝘁𝘂𝗻𝗲 𝘁𝗵𝗲 𝘄𝗵𝗼𝗹𝗲 𝘀𝘆𝘀𝘁𝗲𝗺 and they know which layer to touch first based on whether they’re chasing quality, cost, or latency. Which layer has given you the biggest win so far: agent logic, RAG, inference serving, routing, or GPU ops?
-
Kumaran Ponnambalam shared this𝗪𝗵𝘆 𝗱𝗼 𝗹𝗼𝗻𝗴-𝗰𝗼𝗻𝘁𝗲𝘅𝘁 𝗟𝗟𝗠 𝗱𝗲𝗽𝗹𝗼𝘆𝗺𝗲𝗻𝘁𝘀 𝗵𝗶𝘁 𝗮 𝘄𝗮𝗹𝗹 𝗲𝘃𝗲𝗻 𝘄𝗵𝗲𝗻 𝘆𝗼𝘂 𝗵𝗮𝘃𝗲 𝗽𝗹𝗲𝗻𝘁𝘆 𝗼𝗳 𝗚𝗣𝗨 𝗰𝗼𝗺𝗽𝘂𝘁𝗲? The bottleneck is often the KV cache: it avoids recomputing attention, but it grows with context length and quickly becomes a GPU memory and bandwidth problem A recent paper ( https://lnkd.in/gRqBt6dd ) maps the KV-cache optimization space into five practical levers: 𝗘𝘃𝗶𝗰𝘁: keep only the most useful history (windowing, top-K, learned eviction). 𝗖𝗼𝗺𝗽𝗿𝗲𝘀𝘀: quantize/compress KV tensors to cut footprint with controlled quality tradeoffs. 𝗛𝘆𝗯𝗿𝗶𝗱 𝗺𝗲𝗺𝗼𝗿𝘆: keep hot KV on GPU and page cold KV to CPU/NVMe with smart prefetch. 𝗥𝗲𝘁𝗵𝗶𝗻𝗸 𝗮𝘁𝘁𝗲𝗻𝘁𝗶𝗼𝗻: use mechanisms that reduce dependence on full-history KV. 𝗖𝗼𝗺𝗯𝗶𝗻𝗲: use adaptive pipelines that mix these based on workload and hardware limits. 𝗞𝗲𝘆 𝘁𝗮𝗸𝗲𝗮𝘄𝗮𝘆: there’s no single winner; the right KV strategy depends on your serving mode (multi-turn chat vs long-context analysis vs high-throughput serving), and the best results often come from hybrid, scenario-driven combinations.
-
Kumaran Ponnambalam shared this𝗜𝗳 𝗞𝘂𝗯𝗲𝗿𝗻𝗲𝘁𝗲𝘀 𝘀𝘁𝗮𝗻𝗱𝗮𝗿𝗱𝗶𝘇𝗲𝗱 𝗰𝗼𝗺𝗽𝘂𝘁𝗲 𝗳𝗼𝗿 𝗺𝗶𝗰𝗿𝗼𝘀𝗲𝗿𝘃𝗶𝗰𝗲𝘀, 𝘄𝗵𝗮𝘁’𝘀 𝘁𝗵𝗲 “𝗞𝘂𝗯𝗲𝗿𝗻𝗲𝘁𝗲𝘀” 𝗳𝗼𝗿 𝗱𝗲𝗹𝗶𝘃𝗲𝗿𝗶𝗻𝗴 𝗲𝗻𝘁𝗲𝗿𝗽𝗿𝗶𝘀𝗲 𝗸𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲 𝘁𝗼 𝗔𝗜 𝗮𝗴𝗲𝗻𝘁𝘀 𝘀𝗮𝗳𝗲𝗹𝘆 𝗮𝗻𝗱 𝗰𝗼𝗻𝘀𝗶𝘀𝘁𝗲𝗻𝘁𝗹𝘆? A recent arXiv paper proposes exactly that: Context Kubernetes, a declarative orchestration layer for “enterprise knowledge delivery” in agentic AI systems. https://lnkd.in/ghWkVYwc 𝗖𝗼𝗻𝘁𝗲𝘅𝘁 𝗞𝘂𝗯𝗲𝗿𝗻𝗲𝘁𝗲𝘀 treats knowledge access like a platform problem, not a per-agent RAG script: you declare what knowledge an agent should be able to access, under what permissions, freshness constraints, and governance rules—and a controller reconciles reality to match the desired state. Most enterprises are heading toward “agent sprawl,” where each agent ships its own retrieval logic, permission assumptions, and freshness guarantees—leading to stale context, cross-tenant leakage risk, and inconsistent behavior across teams. The paper’s point is that knowledge for agents needs the same operational rigor we applied to infrastructure. Key implementation constructs (high level) 𝗗𝗲𝗰𝗹𝗮𝗿𝗮𝘁𝗶𝘃𝗲 𝗺𝗮𝗻𝗶𝗳𝗲𝘀𝘁𝘀 (YAML): define “knowledge-architecture-as-code” for what agents can access and how it must be delivered. 𝗥𝗲𝗰𝗼𝗻𝗰𝗶𝗹𝗶𝗮𝘁𝗶𝗼𝗻 𝗹𝗼𝗼𝗽: continuously checks for drift (stale/unauthorized content) and corrects it, Kubernetes-style. 𝗣𝗲𝗿𝗺𝗶𝘀𝘀𝗶𝗼𝗻 𝗺𝗼𝗱𝗲𝗹: a multi-tier authorization design where agent authority is explicitly bounded as a strict subset of human authority. 𝗚𝗼𝘃𝗲𝗿𝗻𝗮𝗻𝗰𝗲 𝗰𝗼𝗻𝘁𝗿𝗼𝗹𝘀: separation of “approval channels” from “knowledge channels,” plus explicit freshness monitoring.
-
Kumaran Ponnambalam shared this𝗗𝗶𝗱 𝗮𝗴𝗲𝗻𝘁𝗶𝗰 𝗔𝗜 𝗷𝘂𝘀𝘁 𝗰𝗵𝗮𝗻𝗴𝗲 𝘆𝗼𝘂𝗿 𝗕𝘂𝗶𝗹𝗱 𝘃𝘀 𝗕𝘂𝘆 𝗱𝗲𝗰𝗶𝘀𝗶𝗼𝗻 𝗺𝗼𝗿𝗲 𝘁𝗵𝗮𝗻 𝗮𝗻𝘆 𝗦𝗮𝗮𝗦 𝘁𝗿𝗲𝗻𝗱 𝗶𝗻 𝘁𝗵𝗲 𝗹𝗮𝘀𝘁 𝗱𝗲𝗰𝗮𝗱𝗲? I put together a simple table that compares build vs buy through the lens of how agentic AI changes the key decision factors . Big takeaway: building is faster than ever, but the center of gravity shifts to operating safely in production—policies, auditability, rollback, and ownership. If you’re evaluating software decisions right now, which factor is most decisive for you: speed, compliance/governance, integration complexity, or long-term ownership?
-
Kumaran Ponnambalam shared this𝗜𝗳 𝗟𝗟𝗠𝘀 𝗮𝗿𝗲 𝘀𝗼 𝗳𝗹𝘂𝗲𝗻𝘁, 𝘄𝗵𝘆 𝗱𝗼 𝘁𝗵𝗲𝘆 𝘀𝘁𝗶𝗹𝗹 𝘀𝘁𝘂𝗺𝗯𝗹𝗲 𝗼𝗻 𝗿𝘂𝗹𝗲-𝗵𝗲𝗮𝘃𝘆 𝗱𝗲𝗰𝗶𝘀𝗶𝗼𝗻𝘀 𝘄𝗵𝗲𝗿𝗲 𝗰𝗼𝗿𝗿𝗲𝗰𝘁𝗻𝗲𝘀𝘀 𝗮𝗻𝗱 𝘁𝗿𝗮𝗰𝗲𝗮𝗯𝗶𝗹𝗶𝘁𝘆 𝗺𝗮𝘁𝘁𝗲𝗿? They fail because they’re optimized for producing plausible text, not executing formal rules: they can miss hidden constraints, "average out" exceptions, struggle to consistently apply multi-step logic, and rarely produce auditable reasoning paths that prove which rule or policy drove a decision. Neurosymbolic AI addresses this by combining neural models (LLMs/NNs) for understanding messy language and data, with symbolic systems (rules, logic, knowledge graphs) for deterministic reasoning, constraints, and verifiable decision trails. https://lnkd.in/gg3knpFc Common architecture patterns for Neurosymbolic AI with LLMs. 𝟭. 𝗟𝗟𝗠 𝗮𝘀 𝗽𝗮𝗿𝘀𝗲𝗿 -> 𝘀𝘆𝗺𝗯𝗼𝗹𝗶𝗰 𝗲𝘅𝗲𝗰𝘂𝘁𝗼𝗿 : A user asks “Are these 12 vendors eligible under our procurement policy?” and the LLM extracts structured facts (vendor type, spend, region, exceptions) while a rules/logic engine deterministically computes eligibility and returns the decision + which rules fired. 𝟮. 𝗟𝗟𝗠 𝗮𝘀 𝗽𝗹𝗮𝗻𝗻𝗲𝗿 -> 𝗰𝗼𝗻𝘀𝘁𝗿𝗮𝗶𝗻𝗲𝗱 𝘁𝗼𝗼𝗹 𝗲𝘅𝗲𝗰𝘂𝘁𝗶𝗼𝗻 : A change-management agent proposes a rollout plan, but every step is validated against hard constraints (maintenance windows, approvals, dependency ordering) and blocked/rewritten if any constraint fails before any tool call executes. 𝟯. 𝗟𝗟𝗠 + 𝗸𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲 𝗴𝗿𝗮𝗽𝗵 𝗿𝗲𝗮𝘀𝗼𝗻𝗶𝗻𝗴 : A support agent answers "Why did customer X’s software fail after release Y?" by traversing a knowledge graph (customer -> services -> incidents -> deployments -> config changes), then uses symbolic path evidence to justify a multi-hop explanation. 𝟰. 𝗣𝗿𝗼𝗴𝗿𝗮𝗺-𝗼𝗳-𝘁𝗵𝗼𝘂𝗴𝗵𝘁 -> 𝗲𝘅𝗲𝗰𝘂𝘁𝗲 𝗱𝗲𝘁𝗲𝗿𝗺𝗶𝗻𝗶𝘀𝘁𝗶𝗰𝗮𝗹𝗹𝘆 : A finance ops assistant converts "reconcile these statements and compute variance drivers" into executable code/queries (SQL/Python), runs them in a sandbox, and returns computed results rather than "reasoning in text."
-
Kumaran Ponnambalam reposted thisKumaran Ponnambalam reposted thisI’ve been deep-diving into the best practices for building Agentic Frameworks lately, and I landed on an absolute goldmine of a session from the Toronto Machine Learning Summit (TMLS). Kumaran Ponnambalam, Principal AI Engineer at Cisco, shared some incredibly raw and practical lessons from deploying 100+ agents in the wild. If you are moving from a simple chatbot to a complex agentic workflow, this is the "missing manual" you’ve been looking for. The Big Takeaway: The "Agent Drift" Problem Most teams focus on Day 1 deployment, but the real challenge is Day 60. Even without changing a single line of your code, your agent’s performance can decay. Cisco identifies 4 silent killers: - Goal Drift: Users start asking questions in ways you never anticipated. - Context Drift: The data sources (like Jira or Slack) shift in format or style. - Reasoning Drift: The underlying LLM model undergoes a "silent update." - Collaboration Drift: The "plumbing" between your sub-agents starts to leak. The Playbook: A Sneak Peek at the Do’s & Don’ts The full list of takeaways is extensive, but here are a few high-impact ones that changed how I think about AgenticOps: ✅ DO: - Implement Continuous Calibration: Run evaluations daily to catch "drifts" before your users do. - Practice Strict Data Pruning: Don’t "flush" raw JSON to your LLM. Prune it to exactly what’s needed to keep reasoning sharp. - Use Versioned APIs: Lock your tools to specific versions so a third-party update doesn't break your workflow. ❌ DON'T: - Set and Forget: An unmonitored agent is a liability. Performance will decay over time. - Skip Benchmarking for "Small" Tweaks: Even a one-sentence change to a prompt can trigger a reasoning cascade. - Ignore the "Tool-Call" Level: Don't just look at the final answer—monitor the steps the agent took to get there. High-five to Kumaran and the TMLS team for sharing this! I’ve only highlighted a few points here, but there is a massive list of Do’s and Don'ts in the full resources. I strongly recommend anyone setting up agentic workflows to check out the original post and the YouTube session for the full deep-dive! 📺👇 https://lnkd.in/gNFXX84YManaging AI Agent Performance Degradation in Production | Kumaran Ponnambalam, CiscoManaging AI Agent Performance Degradation in Production | Kumaran Ponnambalam, Cisco
Kumaran Ponnambalam commented on a post
2mo
Great points Arunabh Bhattacharjee
Kumaran Ponnambalam replied to a comment
5mo
You would have to add automated evaluation against a base dataset to ensure that regression is not lost.
-
Kumaran Ponnambalam liked thisKumaran Ponnambalam liked thisI'm always blown away by the volume and quality of submissions we get for our TMLS - Toronto Machine Learning Summit. I'd like to genuinely thank everyone who took the time to submit 🙏 And as always, we had the impossible task of choosing ~40. Impossible because we undoubtedly turn down many incredible submissions. This is my least favourite part, by far. But one of the most rewarding is working with our committee. They help review and guide decisions. They help identify priority subjects and frame our learnings through countless hours of fun, collaborative discussions. What is left here is a representation of where our current AI ecosystem stands: the wins, the hardships, and the concrete lessons learned.. (and there's a lot of them!). I think the sessions highlighted here are the best we've had over our 10 year run. Each one is intentional, and each one has a strong message for our broader AI ecosystem to take away. If you're interested in joining us this year - shoot me a message or check out the full agenda. And as always, we'll have great discussions, as well as food, drink, and socials. We're at CIBC this year in their beautiful auditorium and conference center. Like last year, space will run out ! ➡️ 🔗 📍 torontomachinelearning.com 🥂 Big thank you to Ozge Yeloglu, Yanmei Zhuo Simon Yau Erin Li and the rest of the CIBC team for all your massive support. TMLS MLOps World | GenAI Summit
-
Kumaran Ponnambalam liked thisKumaran Ponnambalam liked this“Good” security isn’t good enough for Oracle environments anymore. What worked a year or two ago, perimeter defenses and periodic patching, is now leaving critical gaps in an increasingly complex threat landscape. 👉 Join us for a live, executive virtual forum: Securing Oracle Technologies in 2026: ‘Good’ Isn’t Good Enough 📅 June 4 | ⏰ 12–1 PM EST Data Intensity’s Biju Thomas and James White, alongside Brite President Trevor Smith, will share how to better protect Oracle systems—without disrupting always-on operations. You’ll learn: ✅ Where Oracle environments are most exposed ✅ Why traditional controls are falling short ✅ Practical strategies like micro-segmentation, WAFs, and XDR 🔐 Save your seat: https://ow.ly/peXx30sW0s4
-
Kumaran Ponnambalam reacted on thisKumaran Ponnambalam reacted on thisI guess it’s my turn to make this post… Last week, I was part of the layoffs at LinkedIn, alongside so many incredibly talented people. Over the past 6 years, I got to work with countless amazing instructors and wonderful teammates. At the end of every production, we’d always say, “that’s a wrap!” Looks like this time it was for me! 🎬 What a fun chapter. I learned a ton, met incredible people, survived endless calendar invites, and got very good at keeping organized chaos moving. But, the people were always my favorite part. I feel really lucky to have worked with teams and instructors that were smart, creative, funny, and genuinely kind. I’m still my optimistic self and truly feel I’m exactly where I’m supposed to be! So… onto the next adventure! ✨
-
Kumaran Ponnambalam reacted on thisKumaran Ponnambalam reacted on thisThat's a wrap. Five years #LinkedIn. A post in honor of each. This is the first. Grateful to everyone I got to do this purposeful work with. Not everyone gets the chance. Received the #layoff news in a room called Opportunity. That's exactly what this has been. To everyone I've had the privilege to partner with over the years — thank you. If you're building an upskilling culture and looking for someone who finds joy in making learning work, let’s talk. #nextplay 🔗 Quick Links 1 of 5: That's A Wrap https://lnkd.in/gG2dUbMU 2 of 5: Swag For Days → https://lnkd.in/gftSjXud 3 of 5: Claim Your Spotlight → https://lnkd.in/gMtmAYUx 4 of 5: Learn, Build, Share → https://lnkd.in/g5cqnirX 5 of 5: Your Skills. Your Story.→ https://lnkd.in/g2v7bAjs
Experience & Education
-
Cisco
View Kumaran’s full experience
See their title, tenure and more.
or
Licenses & Certifications
-
Deep Learning Specialization
Coursera Course Certificates
-
Data Science Specialization
Coursera Verified Certificates
Publications
-
Want Data Scientists do which traditional Data Analysts dont.
InsideBigData
See publicationComparison of roles and responsibilities between Data Scientists and Data Analysts
-
Statistical Performance Analytics for Agent Scoring
Transera /youtube
See publicationA Webinar on statistical performance analytics.
Recommendations received
12 people have recommended Kumaran
Join now to viewOther similar profiles
Explore more posts
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content