Portfolio — Initialising
SAYAK NASKAR
0%
Loading
AI + QA Automation Engineer · Hyderabad

SAYAKNASKAR.

01Bug hunter.
02System architect.
03AI researcher.
5+ years· AI Researcher· 1 CVE published· OSS contributor
Scroll
Impact — in numbers
0%Production incidents eliminated
0%Manual testing removed from pipeline
$0Client pipeline secured through automation proof
0Security vulnerabilities responsibly disclosed
01
2017 — 2020
The
Hunter.

I spent three years breaking the systems everyone else assumed were safe. Nokia, SAP, Yahoo Paranoid — none of them were.

CVE-2020-8825 Nokia · SAP · Yahoo · Dutch Govt and many more Responsible Disclosure
02
2020 — 2024
The
Architect.

I built the frameworks. The kind that don't just test software — they tell you where the next failure will come from before it happens.

$2M+ pipeline TCS BFSI Scale Selenium · Java · Maven · Junit 1500+ automated Testcases 300+ critical automated scenarios
03
2024 — 2026
The
Leader.

US Government infrastructure. Zero tolerance margins. Teams that I led grew into teams that didn't need me — which is exactly the point.

Government scale QA Strategy Team Leadership Azure · Jenkins · Docker · Log4J
04
2025 — Present
The
Researcher.

M.Tech @ IIIT Guwahati. Two papers under review. Teaching machines the hardest thing in QA: knowing what they don't know.

EAAAI 2026 IEEE TAC 2026 NLP · Transformers · LLMs · SLMs IIIT Guwahati
04Work

What I've
shipped.

01
Government Infrastructure
Public Sector · U.S. State Government

Quality architecture for mission-critical citizen-facing platforms. Built for zero tolerance — because when the system fails, real people don't get paid.

2024–2026
02
Financial Services Architecture
TCS BFSI · Enterprise Scale

Automation framework for a global financial platform. Built to outlast the team that built it — and it did, securing three additional enterprise engagements.

2020–2024
05Research

What I'm
building.

AI Contextual Framework
EAAAI (ex-EANN) 2026 · Core Conf.
CenterDistill: Weakly-Supervised Distillation for Ambiguity-Aware Cross-Lingual QA

A framework that teaches QA models to handle ambiguity — to know when to answer, when to ask, and when to say nothing at all. The hardest problem in cross-lingual question answering.

GitHub ↗ Under Review
Emotion Classification Benchmark
IEEE TAC · Q1 Journal
EMBER: Multi-Seed, Multi-Condition Study of Prompt-Based Emotion Classification

Proving that prompt design, not parameter count, is what determines emotion classification quality. A systematic benchmark across five transformer architectures that changes the question.

Link coming soon ⏳ Under Review
06Recognition

What I've
earned.

CVE-2020-8825
PHP VanillaForum · Responsible Disclosure
Oracle CloudWorld 2023
TCS Partnership · 10,000+ Attendees · Innovation Showcase
2023Global Stage
TCS Client Excellence Award
Client satisfaction 3.2 → 4.7 / 5 · Transformational delivery
2022Excellence Award
Open Source Contributor
Microsoft Powershell · Snapfix (Pypi) · Voice based Email  ·  github.com/hacky1997 ↗
2020–OSS Contributor
07Get in touch

Open to
leadership.

"I spent years finding the breaks.
Now I spend them making sure
the breaks never find you."