Open to AI/ML roles & research positions

Aryan Singh
AI/ML Engineer

B.Tech CSE graduate with research at ISRO Space Applications Centre and IIT Ropar. Built a wake-word detection model for the TRISHNA satellite, a wildfire risk classifier from satellite imagery (F1: 0.914), and shipped a computer vision pipeline that ranked top 0.3% in Flipkart GRID 6.0 among 1,00,000+ participants. HackerRank Gold Medlist in Python and SQL.

Speech & Audio ML Computer Vision Satellite Imagery Embedded AI Deep Learning
View my work Get in touch
About
By the numbers
25%
Accuracy gain over Vosk & Picovoice — TRISHNA satellite wake-word model
Top 0.3%
Flipkart GRID 6.0 — out of 1,00,000+ participants nationwide
95.3%
Accuracy on wildfire risk classifier from satellite imagery (F1: 0.914)
90th
Percentile — GATE 2026, Computer Science & IT (nationally qualified)
<200ms
Real-time inference latency on embedded satellite hardware
2×
User engagement lift via clustering & predictive ML at CodeClause

I'm a Computer Science graduate from Mohanlal Sukhadia University, Udaipur. I focus on the gap between ML research and real-world deployment, building systems that work under constraints, not just in notebooks. I spent months at ISRO's Space Applications Centre engineering a mission-critical speech model for the TRISHNA satellite. My final-year project was a wildfire risk classifier trained on satellite imagery and was submitted as a GSoC 2026 proposal to the University of Alaska Anchorage. I was selected for a nationally competitive 2-month residential AI program at IIT Ropar and am GATE 2026 qualified (90th percentile, CS&IT).

Experience
Where I've worked
ISRO — Space Applications Centre, Ahmedabad
Research Scholar
Jan 2025 – Apr 2025
  • Engineered CNN + MFCC-based wake-word detection model for the TRISHNA satellite's voice command system, achieving 25% higher accuracy than Vosk and Picovoice at 80%+ ambient noise levels.
  • Overcame limited training data using noise augmentation (white noise + time shifting), tripling effective dataset size and significantly improving robustness under real satellite payload conditions.
  • Validated real-time inference on resource-constrained embedded hardware; achieved sub-200ms response latency with fault-tolerant activation logic.
  • Collaborated with 5 ISRO scientists and engineers to integrate the model into the satellite's embedded control system, improving overall voice-command efficiency by 20%. Received Letter of Recommendation from Division Head, PCSVD, SAC.
CNNMFCCTensorFlowLibrosaEmbedded MLNoise Augmentation
IIT Ropar — CSE Dept. / iHub-AWaDH
Research Intern · AI Vicharaṇā Shālā
May 2024 – Jul 2024
  • Selected for nationally competitive 2-month residential AI programme at IIT Ropar (CSE Dept. × iHub-AWaDH, DST-funded AI hub) — one of 30 participants chosen across India.
  • Studied and implemented AI systems across computer vision (object detection, OpenCV), steganography (LSB encoding, Python), and recommendation engines (collaborative filtering, K-means, Plotly 3D visualisation).
  • Gained in-depth exposure to CNN architectures, data augmentation, and model evaluation — knowledge directly applied in the subsequent TRISHNA satellite ML project at ISRO.
OpenCVKerasObject DetectionSteganographyRecommendation SystemsPlotly
CodeClause Pvt. Ltd., Udaipur
Data Science Intern
Dec 2023 – Jan 2024
  • Applied K-means clustering on real user behavioural data to personalise push notification content and timing, resulting in 100% increase in user engagement over a 4-week A/B test cycle.
  • Built an interactive Power BI dashboard analysing e-commerce revenue, customer behaviour, product category trends, and regional performance — visualising 500K+ transactions with 8+ charts and DAX measures.
K-meansScikit-learnPower BIDAXA/B Testing
Projects
Things I've built
🛰
TRISHNA Satellite Wake-Word System
CNN + MFCC binary wake-word classifier for ISRO's TRISHNA satellite voice command system. Noise augmentation pipeline (white noise + time shift). Outperformed Vosk and Picovoice by 25% at 80%+ ambient noise. Sub-200ms inference on embedded hardware.
TensorFlow Librosa MFCC CNN Embedded
View on GitHub →
🔥
Wildfire Risk Prediction GSoC 2026
CNN classifier trained on 2,000+ labelled satellite images of boreal forests — 95.3% accuracy, 94% precision, 96% recall, F1: 0.914. Deployed as a live Streamlit web app that accepts satellite image uploads and returns fire/no-fire probability in real time. Submitted as GSoC 2026 proposal.
TensorFlow OpenCV Streamlit Satellite Data CNN
View on GitHub →
📦
Smart Vision Quality Control
End-to-end CV quality control pipeline — freshness detection (Keras CNN), product recognition (YOLOv8, 5,000+ images), OCR label extraction (Tesseract), and item counting. Deployed as a multi-page Streamlit app with SQLite logging. Ranked top 0.3% in Flipkart GRID 6.0.
YOLOv8 Roboflow Streamlit SQLite OpenCV
View on GitHub →
🔐
Image · Audio · Emoji Steganography
Three steganography techniques in one toolkit — LSB encoding in images, LSB encoding in audio files, and zero-width Unicode character encoding in emoji text. Built during IIT Ropar AI programme.
Python OpenCV NumPy wave
View on GitHub →
📖
OCR Speaking Lamp
Raspberry Pi accessibility device that reads printed text aloud in real time. Tesseract OCR + pyttsx3 text-to-speech pipeline. 95% OCR accuracy — fully offline, designed for visually impaired users. No internet required.
Raspberry Pi Tesseract OCR pyttsx3 OpenCV
View on GitHub →
🛍
Retail Recommendation System
Collaborative filtering engine (user-based + item-based) on 500K+ retail transactions. Item-based CF achieves Precision@10: 0.34, covering 52% of product catalogue. Interactive 3D Plotly visualisation for cluster analysis.
Scikit-learn Pandas Plotly K-means
View on GitHub →
Skills
What I work with
Languages
Python C++ Java SQL
ML / AI Frameworks
PyTorch TensorFlow Scikit-learn OpenCV Keras Librosa Pytesseract Seaborn
Specialisations
Speech Processing Computer Vision Embedded ML Edge AI Satellite Imagery Recommendation Engines OCR NLP
Tools & Platforms
Streamlit Git / GitHub Raspberry Pi Roboflow Power BI Google Colab MySQL SQLite Jupyter
Recognition
Achievements & certifications
📊
GATE 2026 Qualified — 90th Percentile, CS & IT
Nationally qualified in Computer Science & Information Technology
🧪
Letter of Recommendation — Division Head, ISRO SAC
Issued by Head of PCSVD, Space Applications Centre, Ahmedabad — for outstanding research on the TRISHNA satellite voice command project
🏆
Top 0.3% — Flipkart GRID 6.0 (1,00,000+ participants)
Software Development, Information Security & Robotics tracks — Certificate of Merit, Flipkart · Semi-finalist, Robotics Track
🌍
GSoC 2026 Proposal — Wildfire Risk Prediction
Submitted to University of Alaska Anchorage — CNN satellite imagery classifier, F1: 0.914, 95.3% accuracy · Live Streamlit demo
🥇
HackerRank Gold Medal — Python & SQL
Verified skill certifications: Python (Basic) — Feb 2024 · SQL (Basic) — May 2024
💼
Graduate Engineer Trainee offer — Secure Meters Ltd.
Highest placement offer from the 2025 graduating batch at IET, MLSU Udaipur
🎓
IIT Ropar — AI Vicharaṇā Shālā Completion Certificate
Nationally selected · 2-month residential hands-on AI programme · CSE Dept. × iHub-AWaDH · Grade A+

Let's work together

Open to AI/ML engineering roles, research positions, and collaborative projects. Based in Udaipur — open to inoffice, and remote jobs.

aryansingh4653@gmail.com LinkedIn GitHub Resume PDF