---
title: "Aditya Kumar - Genai / Ml engineer in Delhi | Tal"
description: "Meet Aditya Kumar, Analyst Programmer L2 at Xceedance based in Delhi. AI stack includes antigravity, claude, and gemini."
image: "https://tal.club/meet/aditya-kumar-36/share-image"
---

# Aditya Kumar on Tal

## Headline

Analyst Programmer L2 at Xceedance

## About

I’m a Data & AI Engineer with 4\+ years of experience building scalable data platforms, machine learning solutions, and production\-grade GenAI applications\. My experience spans Data Engineering, Generative AI, LLMs, Computer Vision, and Backend Engineering, with a strong focus on turning complex business problems into reliable, scalable solutions\. Currently, I work as a Python Developer L2 in an R&D team, where I build AI\-powered systems using LLMs, RAG, prompt engineering, embeddings, vector databases, FastAPI, and cloud technologies\. I’ve worked on production GenAI pipelines for document understanding and structured data extraction, including multi\-stage LLM workflows for processing insurance documents\. I’ve also built computer vision systems for aerial imagery, automated PDF/table extraction pipelines, and AI\-powered address parsing solutions\. Previously, as a Data Engineer at Celebal Technologies, I designed and optimized end\-to\-end data pipelines using Azure, Databricks, ADF, PySpark, PostgreSQL, and Delta Lake\. I led engineering initiatives, migrated large\-scale workloads to Azure, built streaming and batch pipelines, and optimized legacy workloads to achieve significant improvements in execution time and data\-processing efficiency\. Core areas I work in: • Generative AI & LLM Applications • RAG, Prompt Engineering & LLM Fine\-Tuning • Computer Vision & Deep Learning • Data Engineering & ETL/ELT • PySpark, Databricks & Apache Airflow • FastAPI & Microservices • Azure & AWS • PostgreSQL, MongoDB, Redis & Vector Databases • Python, SQL & JavaScript I’m particularly interested in GenAI, AI agents, LLM systems, scalable data architectures, and applied machine learning, and I’m always looking for opportunities to build products that solve meaningful real\-world problems\. engineering and cybersecurity\.

## Location

Delhi

## AI stack

antigravity

claude

gemini

hugging face

langchain

langgraph

## Work experience

Genai / Ml engineer at Xceedance, May 2025 \- Present · 1 y 3 mo, Noida

Python Developer at Xceedance Consulting India Private Limited, May 2025 \- Present · 1 yr 4 mos, Built and operated a multi\-stage LLM extraction pipeline that ingests scanned/PDF insurance placement slips and emits structured policy data \(parties, limits, sublimits, deductibles, signed\-line underwriters\) into MRF Excel templates\. Designed a two\-pass GenAI flow: \(1\) vision\-LLM PDF → Markdown conversion preserving layout for tables and signed lines; \(2\) schema\-driven field extraction with optional structured outputs and a re\-extraction loop for low\-confidence fields, validated via Pydantic models and post\-processed into zone\-aware DataFrames\. Architected decoupled FastAPI microservices \(Job Scheduler \+ GenAI Workers\) with REST \+ JWT auth, CSRF, CORS, and job\-completion callbacks\. Designed an LLM token\-bucket coordinator with DRR fair scheduling and per\-key rate limits \(allocate / deallocate flow\)\. Wired MongoDB \(Cosmos\) \+ SQLite dual storage, Azure Key Vault for key rotation, and Azure File Share for I/O\. Led end\-to\-end CV pipeline for aerial roofs, delivering a 5\-class roof cover classifier \(ResNet152\) and roof geometry detector using TensorFlow and OpenCV\. Engineered SIFT\-enhanced preprocessing and advanced augmentation strategies, improving top\-1 accuracy by ∼8–12% and recall on noisy tiles by ∼15–20%\. Built and deployed a FastAPI \+ TensorFlow Serving microservice containerized with Docker, reducing P95 latency to < 120 ms and supporting > 5k req/min throughput\. Implemented automated training workflows \(checkpointing, LR scheduling, regularization\), boosting macro\-F1 by 5–7% and reducing inference cost by ∼25–30% via batch optimizations\. Built a Python borderless\-table detector for PDFs using pdfplumber word\-bbox extraction to mask text, OpenCV Canny \+ line\-segment clustering/merging to recover horizontal rules, then segmented row bands between consecutive horizontals and exported structured output to Excel \(openpyxl\) with QA overlay PDFs \(pypdf \+ ReportLab\)\. Developed and deployed an AI\-powered address parser using GPT\-4\.

Associate \- Big Data at Celebal Technologies, Jan 2022 \- Present · 4 yrs 8 mos, Noida, Uttar Pradesh, India

Data Engineer at Celebal Technologies, Apr 2022 \- May 2025 · 3 yrs 2 mos, Leading the development of nudges and writing analytical scripts, creating API’s which is used to send targeted WhatsApp messages to the field force\. Developed end to end automated solutions by creating pipelines and scripts along with email notifications using Logic Apps, encompassing data ingestion, Databricks scripts, and result delivery via email to various stakeholders directly\. Contributed to implementation of ADF from its nascent stage, designing and building the current architecture\. Specialized in data wrangling and converting legacy pandas code to PySpark, resulting in more than 80% of reduction in execution time\. Led a team of 3 engineers in designing and building robust end\-to\-end data pipelines for both streaming and batch migration of PostgreSQL database to blob storage leveraging Databricks, ADF and IOTHub\. Successfully migrated a large\-scale PostgreSQL database to Azure Storage, improved data storage and retrieval efficiency by 30% through advanced techniques and code optimizations\. Extracted the customer data from APIs in JSON format to get to know the exact location of their ships using latitude and longitude\. Implementing CDC logic and SCD logic to track the changes\. Processing real time streaming event\-based data to get the status of attached devices with the ship\. Monitored and Optimized devices Performance Including latency and throughput\. Migrated the entire existing ETL workload from SAP to Azure Cloud \(using ADF, Databricks, etc\.\), including all queries, tables, stored procedures, and functions\. Responsible for designing and developing the ETL pipelines using Azure tools\. Used Python Scripting for automation of converting simple queries to PySpark, accelerating the process significantly\. Developed scripts to perform schema validation and data consistency checks\. Validated end\-to\-end workflows to ensure no data loss or corruption\. Perform regression testing to confirm the functionality of migrated workloads\.

Cyber Security Specialist at aNumak & Company, Aug 2021 \- Sep 2021 · 2 mos, India

Management Trainee at Homeflic Wegrow, Mar 2021 \- Jun 2021 · 4 mos, India

## Education

Bachelor of Technology \- BTech, Computer/Information Technology Administration and Management, Swami Keshwanand Inst\. Of Tech\. Mgt\. & Gramothan,Jaipur

Bachelor of Technology, Computational Science, SKIT Jaipur

[View the original profile](https://tal.club/meet/aditya-kumar-36)

```json
{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "Person",
      "@id": "https://tal.club/meet/aditya-kumar-36#person",
      "name": "Aditya Kumar",
      "url": "https://tal.club/meet/aditya-kumar-36",
      "image": "https://tal.club/meet/aditya-kumar-36/profile-photo.jpg",
      "description": "Meet Aditya Kumar, Analyst Programmer L2 at Xceedance based in Delhi. AI stack includes antigravity, claude, and gemini.",
      "jobTitle": "Genai / Ml engineer",
      "worksFor": {
        "@type": "Organization",
        "name": "Xceedance"
      },
      "hasOccupation": {
        "@type": "Occupation",
        "name": "Genai / Ml engineer"
      },
      "homeLocation": {
        "@type": "Place",
        "name": "Delhi"
      },
      "alumniOf": [
        {
          "@type": "EducationalOrganization",
          "name": "Swami Keshwanand Inst. Of Tech. Mgt. & Gramothan,Jaipur"
        },
        {
          "@type": "EducationalOrganization",
          "name": "SKIT Jaipur"
        }
      ],
      "knowsAbout": [
        "antigravity",
        "claude",
        "gemini",
        "hugging face",
        "langchain",
        "langgraph"
      ],
      "mainEntityOfPage": {
        "@id": "https://tal.club/meet/aditya-kumar-36#webpage"
      }
    },
    {
      "@type": "ProfilePage",
      "@id": "https://tal.club/meet/aditya-kumar-36#webpage",
      "url": "https://tal.club/meet/aditya-kumar-36",
      "name": "Aditya Kumar on Tal",
      "description": "Meet Aditya Kumar, Analyst Programmer L2 at Xceedance based in Delhi. AI stack includes antigravity, claude, and gemini.",
      "dateModified": "2026-08-13T09:12:34.875Z",
      "isPartOf": {
        "@id": "https://tal.club/#website"
      },
      "mainEntity": {
        "@id": "https://tal.club/meet/aditya-kumar-36#person"
      },
      "primaryImageOfPage": {
        "@type": "ImageObject",
        "url": "https://tal.club/meet/aditya-kumar-36/profile-photo.jpg",
        "contentUrl": "https://tal.club/meet/aditya-kumar-36/profile-photo.jpg",
        "caption": "Aditya Kumar profile photo on Tal"
      },
      "inLanguage": "en-IN"
    }
  ]
}
```