Skip to content
SSM

Shoaib Sadiq Salehmohamed

OBSERVATORY · 17.3850°N 78.4867°E · HYDERABAD

I build AI systems that turn uncertainty into usable signal.

A research intern working on LLM reliability and hallucination detection, a Smart India Hackathon 2025 winner, and a full-stack builder shipping products across AI, threat intelligence, robotics, and developer tools.

View selected workRead researchRésumé ↗

Currently: LLM reliability · hallucination detection · threat intelligence · full-stack AI products

signal // live
SSM·2026
noisesignal
  • Smart India Hackathon 2025 — WinnerDec 2025
  • Research Intern — LLM Lens2026—
  • Patent published — RakshakJan 2026
  • OpenAI ChatGPT Lab — Student cohort2025
  • arXiv:2604.06277Paper
CH·01/Selected proof
SECTOR: PROOF · CLASS: PUBLIC

Not a list of skills — a short stack of things with proof behind them.

01National Win

Smart India Hackathon 2025 Winner

Built TattvaDrishti, an AI threat-intelligence platform for LLM-powered misinformation.

Open →
02Research

LLM Reliability Research

Co-authored arXiv:2604.06277 on hidden-state hallucination detection at LLM Lens.

Open →
03Patent

Patent-published robotics

Co-inventor on Rakshak, an autonomous surveillance rover (App. No. 202541112724 A).

Open →
04Cohort

OpenAI ChatGPT Lab

Selected among 28 students nationwide for India's inaugural student cohort.

Open →
CH·02/Core direction
SECTOR: DIRECTION · 3 CHANNELS

I work where AI research meets shipped product.

01

Reliable AI

Hallucination detection, uncertainty estimation, hidden-state probes, evaluation pipelines, and model-behavior analysis.

02

Intelligence Systems

Threat intelligence, misinformation analysis, graph-based risk scoring, and analyst-facing dashboards.

03

Full-Stack Execution

Next.js, FastAPI, PyTorch, real-time interfaces, and data pipelines — products built under hackathon-speed constraints.

CH·03/Featured work
SECTOR: ARCHIVE · ENTRIES: 05

Five things that matter — each one a problem turned into a working system.

All work →
Entry 01SIH 2025 Winner

TattvaDrishti

AI threat-intelligence platform

An end-to-end platform that detects, attributes, and explains LLM-powered misinformation for analysts.

Content
Detection
Attribution
Risk score
Analyst dashboard
yr 2025FastAPINext.jsPyTorch
Open dossier→
Entry 02arXiv Paper

LLM Reliability Research

Hallucination detection · LLM Lens

Detecting hallucinations from inside the model — reading hidden states before unreliable text reaches a user.

Prompt
Model
Hidden states
Probe
Reliability signal
yr 2025–presentPyTorchTransformersHugging Face
Open dossier→
Entry 03Patent Published

Rakshak

Autonomous surveillance rover

A real-world autonomous rover for perimeter surveillance — perception, localization, and safety holding together in the field.

Sensors
Perception
Localization
Control
Autonomous patrol
yr 2026Jetson NanoPixhawkLiDAR
Open dossier→
Entry 04Full-stack AI

LingualLink

Real-time multilingual chat

A real-time chat platform that translates across languages live, built on a production-shaped stack.

Message
Inference
Translation
WebSocket
Delivery
yr 2025FastAPINext.jsWebSockets
Open dossier→
Entry 05Active Prototype

AstraOS

Memory-augmented AI workspace

A privacy-first personal AI workspace with persistent memory and document-grounded retrieval.

Query
Retrieval
Memory
Local LLM
Response
yr 2025–presentFastAPIReactTypeScript
Open dossier→
CH·04/Research
SECTOR: RESEARCH · CLASS: ONGOING

Researching how language models fail — and how systems can respond.

At LLM Lens I study whether hidden states, cross-layer behavior, and uncertainty signals can detect unreliable generations before they reach a user — reading reliability from inside the model rather than checking it after the fact.

detection path
Prompt
Model
Hidden states
Probe
Reliability signal

Hidden-state probing

Whether internal activations encode hallucination signals across layers and datasets — and how to read them.

Cross-model portability

Whether reliability detectors trained on one model family transfer to another, or have to be re-learned.

Reliability controllers

Lightweight control layers that can detect, abstain, or seek evidence during generation, before text reaches a user.

Read research notes →
CH·05/Build log
SECTOR: LOG · CHRONOLOGICAL

A running record of what got built, won, and shipped.

  1. 2026

    Research Intern — LLM Lens

    LLM reliability, hallucination detection, and hidden-state evaluation pipelines.

  2. 2026

    Rakshak patent published

    Co-inventor on an autonomous surveillance rover — Indian Patent App. No. 202541112724 A.

  3. 2025

    Smart India Hackathon Winner

    Built TattvaDrishti for AI-driven threat intelligence under the AICTE Cybersecurity Cell.

  4. 2025

    OpenAI ChatGPT Lab for Students

    Selected for India's inaugural student cohort (28 nationwide).

  5. 2024

    Dual academic track begins

    B.Tech CSE at KMIT alongside the IIT Madras BS in Data Science & Applications.

  6. 2024

    Open source — GSSoC + Hacktoberfest

    Top 8% of GSSoC contributors; Hacktoberfest Super Contributor.

CH·06/Toolchain
SECTOR: SYSTEMS

What I reach for.

AI / ML
PyTorchTransformersHugging FaceOpenCVPandasNumPyMatplotlib
Backend
FastAPIFlaskNode.jsExpressRedisCelerySQLAlchemy
Frontend
Next.jsReactVueTailwindFramer Motion
Systems
DockerLinuxWebSocketsGitHub ActionsAzureAWSGCP
Languages
PythonJavaCJavaScriptTypeScriptSQL
CH·07/Contact
SECTOR: COMMS · CHANNEL: OPEN

Want to build, research, or collaborate?

I’m open to research collaborations, AI / product internships, hackathon teams, and ambitious full-stack builds.

ResumeDownload PDF↗Emailshoaibkulsums@gmail.com↗GitHubgithub.com/Shoaib2000857↗LinkedInlinkedin.com/in/shoaib-ssm↗YouTube@shoaibsadiq2238↗