---
title: "Aditi Singla - Systems engineer in India | Tal"
description: "Meet Aditi Singla, ex-Researcher@TCS research and innovation | CSE @ IIIT-Delhi (2026) | Applied AI Researcher | LLM Agents, MCTS, Distributed Training (DeepSpeed) | ICLR’24 Author based in India. AI stack includes claude code, codex, and cursor."
image: "https://tal.club/meet/aditi-singla/share-image"
---

# Aditi Singla on Tal

## Headline

ex\-Researcher@TCS research and innovation \| CSE @ IIIT\-Delhi \(2026\) \| Applied AI Researcher \| LLM Agents, MCTS, Distributed Training \(DeepSpeed\) \| ICLR’24 Author

## About

M\.Tech Computer Science student at IIIT\-Delhi, graduating May 2026, actively seeking full\-time roles in Machine Learning Engineering, Applied AI Research, and Software/Backend Engineering\. I build systems that make large language models reason like engineers — not just autocomplete code, but navigate repositories, use tools, and self\-correct\. My thesis designs tool\-augmented agentic pipelines for repository\-level code generation\. I've scaled trajectory collection past 100K samples using beam search and Monte Carlo Tree Search \(MCTS\), turning execution\-grounded correctness into reward signals a model can actually learn from\. Earlier, I spent a semester distributed\-pretraining CodeT5\+ on a 5M\-sample multilingual code corpus with DeepSpeed ZeRO\-2 across 16 GPUs on the Param Rudra HPC cluster — cutting training loss by 54%, and with LoRA fine\-tuning, perplexity by 52% while touching under 3% of the model's parameters\. I'm not only a researcher\. As a backend developer at IIITD, I designed a Django\-based system that took graduation\-eligibility verification from three months to one week, with role\-based REST APIs and real\-time monitoring built in from day one\. My first\-author paper on ensemble methods for language models was accepted at ICLR 2024\. What I care about: agentic AI systems, LLM post\-training, distributed training infrastructure, and backend systems that don't fall over\. Always happy to talk agents, RL\-flavored training signals, or GPU cluster war stories — reach out\.

## Location

India

## AI stack

claude code

codex

cursor

github copilot

hermes agent

langchain

## Work experience

Systems engineer at TCS Research, Jul 2026 \- Oct 2026 · 3 mo, Chennai, Built Linage, a semantic migration pipeline that parses legacy COBOL/C/Java into OMG KDM models and mines provenance\-grounded evidence \(file \+ line citations\) for every business\-rule candidate\. Designed a “propose\-then\-verify” trust boundary: LLM\-proposed contracts are validated by auto\-generated tests executed against the real legacy system, raising rule precision from 71% \(raw LLM\) to 97%\. Implemented differential testing \(legacy vs\. generated target, byte\-identical outputs\) achieving 2,712/2,712 passing cases at \~10K LOC scale, with parsing of a 56K\-LOC codebase in under 2s \(warm cache\)\. Engineered SQLite\-backed semantic memory for incremental re\-migration, reusing 1,252 verified contracts on code changes and running \~5\.8× faster than full re\-migration\. Added polyglot rule reconciliation across COBOL/Java twins, a human\-in\-the\-loop review queue with audit trail, and prompt\-injection/schema\-validation hardening against LLM hallucination\. Built a Flask web dashboard to launch migrations, browse extracted contracts, and drive human review\.

Back End Developer at Graduation Checklist Portal, Jan 2025 \- Jun 2025 · 6 mos, Built and maintained the backend of a Django\-based graduation checklist portal, coding graduation requirement logic into a configurable rules engine that could be updated as academic policy evolved\. Cut graduation\-eligibility processing time from three months to one week by replacing manual transcript checks with an automated, ORM\-driven data pipeline for retrieving and validating student grades and course records\. Reduced the graduation verification workload from a 3–4 person manual process to a single\-person, system\-assisted workflow, improving both speed and accuracy of eligibility assessment\. Integrated automated testing pipelines into the development workflow to catch regressions early and keep the rules engine robust as requirements changed\.

## Education

Master of Technology \- MTech, Computer Software Engineering, Indraprastha Institute of Information Technology, Delhi

Bachelor of Technology \- BTech, Computer Science and Social science Engineering, Indraprastha Institute of Information Technology, Delhi

[View the original profile](https://tal.club/meet/aditi-singla)

```json
{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "Person",
      "@id": "https://tal.club/meet/aditi-singla#person",
      "name": "Aditi Singla",
      "url": "https://tal.club/meet/aditi-singla",
      "image": "https://tal.club/meet/aditi-singla/profile-photo.jpg",
      "description": "Meet Aditi Singla, ex-Researcher@TCS research and innovation | CSE @ IIIT-Delhi (2026) | Applied AI Researcher | LLM Agents, MCTS, Distributed Training (DeepSpeed) | ICLR’24 Author based in India. AI stack includes claude code, codex, and cursor.",
      "jobTitle": "Systems engineer",
      "worksFor": {
        "@type": "Organization",
        "name": "TCS Research"
      },
      "hasOccupation": {
        "@type": "Occupation",
        "name": "Systems engineer"
      },
      "homeLocation": {
        "@type": "Place",
        "name": "India"
      },
      "alumniOf": [
        {
          "@type": "EducationalOrganization",
          "name": "Indraprastha Institute of Information Technology, Delhi"
        },
        {
          "@type": "EducationalOrganization",
          "name": "Indraprastha Institute of Information Technology, Delhi"
        }
      ],
      "knowsAbout": [
        "claude code",
        "codex",
        "cursor",
        "github copilot",
        "hermes agent",
        "langchain"
      ],
      "mainEntityOfPage": {
        "@id": "https://tal.club/meet/aditi-singla#webpage"
      }
    },
    {
      "@type": "ProfilePage",
      "@id": "https://tal.club/meet/aditi-singla#webpage",
      "url": "https://tal.club/meet/aditi-singla",
      "name": "Aditi Singla on Tal",
      "description": "Meet Aditi Singla, ex-Researcher@TCS research and innovation | CSE @ IIIT-Delhi (2026) | Applied AI Researcher | LLM Agents, MCTS, Distributed Training (DeepSpeed) | ICLR’24 Author based in India. AI stack includes claude code, codex, and cursor.",
      "dateModified": "2026-11-04T17:03:58.835Z",
      "isPartOf": {
        "@id": "https://tal.club/#website"
      },
      "mainEntity": {
        "@id": "https://tal.club/meet/aditi-singla#person"
      },
      "primaryImageOfPage": {
        "@type": "ImageObject",
        "url": "https://tal.club/meet/aditi-singla/profile-photo.jpg",
        "contentUrl": "https://tal.club/meet/aditi-singla/profile-photo.jpg",
        "caption": "Aditi Singla profile photo on Tal"
      },
      "inLanguage": "en-IN"
    }
  ]
}
```