Suraj Singh — portfolio with work, education, projects, travel, and tech news

Senior Artificial Intelligence Engineer · Gurugram, India

Suraj Singh AI Engineer.

I design and ship AI-native products end to end — multi-agent LLM systems, retrieval-augmented generation pipelines, production-grade backends, and rigorous evaluation frameworks for LLMs. I turn complex AI capabilities into reliable, scalable software.

Models · Stack · Infra
🐍Python
⚡FastAPI
🔗LangChain
📇LlamaIndex
💬ChatGPT
✶Claude
✦Gemini
🦙Ollama
🤖HF Transformers
🧠vLLM · OpenAI API
☁AWS
🐳Docker
◎Milvus
🔎Solr · Elastic
🗄MySQL · Redis
✈Airflow
🐍Python
⚡FastAPI
🔗LangChain
📇LlamaIndex
💬ChatGPT
✶Claude
✦Gemini
🦙Ollama
🤖HF Transformers
🧠vLLM · OpenAI API
☁AWS
🐳Docker
◎Milvus
🔎Solr · Elastic
🗄MySQL · Redis
✈Airflow

Publications

Fusion Under Component Failure: Negative Results and Failure Modes in Ensemble AI-Generated Image Detection
Singh, S., Verma, T., Singh, P., Bhav, S., Kumar, S. · 2026
arXiv:2609.31749 ↗
SOAR: Advancements in Small Body Object Detection for Aerial Imagery Using State Space Models and Programmable Gradients
Singh, S., Verma, T., Singh, J., Bhartari, Y., Jarwal, R., Singh, S. · 2024
arXiv:2405.01699 ↗

Projects

RAG · NSUT
Retrieval-augmented generation chat interface with LLMs
Built and deployed a chat interface for seamless LLM interactions using the ChainLit framework, with Retrieval-Augmented Generation over a knowledge document. Implemented a knowledge database with LangChain for efficient context retrieval and optimized resource use for foundation-model and OpenAI API inference.
ChainLitLangChainRAGOpenAI
Aug 2023 – Oct 2023↗
Healthcare AI · NSUT
AI for healthcare — diabetic retinopathy
Applied deep learning and image processing to classify fundus images into proliferative and non-proliferative diabetic retinopathy for early diagnosis. Reviewed relevant literature and deployed CNNs and prebuilt architectures (Inception-V3) with transfer learning in the initial phase; explored self-supervised and generative models for later phases.
CNNInception-V3Transfer Learning
May 2023 – Sep 2023↗
NLP · NSUT
MCQ generation with NLP
Streamlined MCQ generation with the T5 transformer and enhanced options using WordNet, ConceptNet, and sense2vec. Extracted text from images with PyTesseract for summarization and used TextRank + RAKE for keyphrase extraction — producing suitable, essential answer choices from extracted information.
T5WordNetConceptNetsense2vec
Feb 2023 – Apr 2023↗
NLP · NSUT
YouTube video summarization & Q&A
Designed a YouTube summarization system leveraging NLP — Pytube to download transcripts and video, BERT extractive summarization to generate concise summaries from transcripts, and MoviePy for video post-processing to enhance the visual presentation. LangChain + LLMs answer user queries about the content.
BERTPytubeMoviePyLangChain
Dec 2022 – Jan 2023↗
NLP · NSUT
Spam classifier with Naive Bayes
Implemented a spam email classifier using the Naive Bayes algorithm with scikit-learn and NLTK for preprocessing, feature extraction, and training. Engineered word-frequency analysis and TF-IDF vectorization; evaluated rigorously with cross-validation and precision / recall / F1.
Naive Bayesscikit-learnNLTKTF-IDF
Oct 2022↗
Tool · NSUT
Voice assistant
Built a Python voice assistant using the Speech Recognition library, allowing users to control devices and perform tasks through voice commands. Used NLP techniques to understand and respond to spoken requests; integrated with external services like Google Search and Amazon Alexa, with unit tests covering reliability and accuracy.
PythonSpeechRecognitionNLP
May 2022 – Jul 2022↗
// world explorer
My Travel Diaries
Stories, photos, and routes from across the globe
8
Countries
24
Cities
∞
Memories
🗼
Paris, France
+
upload photo
🌸
Kyoto
+
upload
🏰
Prague
+
🏝
Bali
+
🕌
Istanbul
+
upload
🎑
Rajasthan
+
upload
🇯🇵
Japan Highlights
7 days
Days 1–3 · Tokyo — Shibuya, Akihabara, TeamLab
Days 4–5 · Kyoto — Fushimi Inari, Arashiyama
Days 6–7 · Osaka — Dotonbori, Nara day trip
🇮🇩
Bali Essentials
5 days
Days 1–2 · Ubud — rice terraces, Monkey Forest
Days 3–4 · Seminyak — beach, cafés, sunsets
Day 5 · Uluwatu — clifftop temple, Kecak dance
🇫🇷
Paris in 4 Days
4 days
Day 1 · Eiffel Tower, Montmartre, Seine cruise
Day 2 · Louvre, Tuileries, Le Marais district
Days 3–4 · Versailles + hidden bookshops
🏔
Patagonia
ARG / Chile
TOP PRIORITY
🌋
Iceland
Northern Lights
AURORA
🦁
Kenya
Safari & Wildlife
BUCKET LIST
🗽
New York
United States
CLASSIC
🛖
Maldives
Indian Ocean
SOMEDAY
🇵🇹
Lisbon
Portugal
SOON
🏯
Rajasthan
India
CLOSE HOME
🎌
Seoul
South Korea
NEXT UP
// staying current
Tech news
Live · May 2026
ARTIFICIAL INTELLIGENCE
THE VERGE · 2 HOURS AGO
Next-gen models are redefining what a developer can ship in a single day
Frontier models now handle end-to-end coding, design reviews, and test generation — fundamentally changing how teams think about sprint velocity and what's possible solo.
The Verge8 min read
HARDWARE
ARS TECHNICA · 5 HOURS AGO
On-device AI chips have crossed the threshold — cloud is now optional for most tasks
Apple, Qualcomm, and MediaTek's latest silicon runs 7B parameter models locally at real-time speed. Privacy-first AI is no longer a performance compromise.
Ars Technica5 min read
STARTUPS
TECHCRUNCH · 1 DAY AGO
YC W25 is the most AI-dense batch in the accelerator's history — 70% AI-native
Investors say the bar for non-AI startups has never been higher to break through. Three companies from the batch are already at $1M ARR within 60 days of demo day.
TechCrunch4 min read
SECURITY
WIRED · 3 HOURS AGO
AI-assisted phishing attacks surge — traditional email security is no longer sufficient
Enterprise security teams are scrambling as LLM-crafted phishing emails achieve near-perfect personalization, bypassing filters trained on older, clunkier attack patterns.
Wired3 min read
// TRENDING NOW
01GPT-5 release82k
02Apple Intelligence 2.041k
03Humanoid robots29k
04EU AI Act compliance17k
05Open-source LLMs14k
06India SaaS $30B ARR9k
🧠
AI daily digest
Get a personalized summary of today's top 5 tech stories, curated to your interests.