ASHOK

Deep learning · Data Analysis · Computer vision

B.Tech — Artificial Intelligence and Data Science
Kumaraguru College of Technology · Coimbatore, Tamil Nadu · 2023 – 2027 · Present
Education
Data Analyst Intern
Roche · Chennai, Tamil Nadu · July 2026 – Present
Internship
  • Conduct structured desk research and market analysis across epidemiology, competitor activity, market access, and customer insights.
  • Consolidate and validate information from multiple sources into comprehensive market insights reports.
  • Perform source validation, deduplication, and gap analysis to improve the accuracy and completeness of market research deliverables.
Innovation Engineer Trainee
Forge Innovation & Ventures · Coimbatore, Tamil Nadu · January 2026 – July 2026
Experience
  • Developed a real-time computer vision system for detecting and classifying Indian traffic signboards using a two-stage YOLO-based pipeline.
  • Improved small and distant sign detection using tiled inference and object tracking, achieving 85.96% mAP@50 and 95.5% classification accuracy.
  • Built the real-time inference pipeline, achieving 15+ FPS in a portable Raspberry Pi 5 CPU deployment unit.
SRFP Research Intern
IIT Gandhinagar · May 2025 – July 2025
Internship
  • Implemented an image enhancement system on Zynq UltraScale+ FPGAs via Vitis-AI.
  • Deployed a 10K-parameter custom finetuned ZeroDCE++ model for real-time inference.
  • Replaced unsupported operators to avoid CPU fallback and reduced complexity by 87.5%.
SafeSense
SafeSense
Compact skeleton and IMU activity recognition designed for robust inference under sensor degradation.
PyTorchSensor FusionActivity Recognition
DeepSeek-OCR Visualized
DeepSeek-OCR Visualized
Live dashboard visualizing OCR inference stages, activations, attention, hidden states, and generated tokens.
ReactWebSocketsThree.js
NanoDCE
NanoDCE
Knowledge distillation compressing ZeroDCE++ down to a tiny model while preserving low-light enhancement quality.
PyTorchComputer VisionKnowledge Distillation
Satellite Visual Search
Satellite Visual Search
Few-shot Siamese + ResNet pipeline to detect objects in satellite imagery with minimal examples.
PyTorchResNet50Siamese Networks
CHB-MIT EEG Analysis
CHB-MIT EEG Analysis
Exploratory analysis comparing seizure vs non-seizure signals with preprocessing and statistics.
PythonEDASignals
FPGA ZeroDCE++
FPGA ZeroDCE++
Deployed and optimized an image enhancement pipeline on DPU architecture of Zynq UltraScale+ zcu104 board via Vitis-AI Framework.
FPGAVitis-AIEdge AI
MuViz
MuViz
Music visualizer with Milkdrop and spectrum modes, queue playback, and synced karaoke-style lyrics.
DjangoJavaScriptWeb Audio
465 / 6125
Kaggle — ROGII Wellbore Geology Prediction
Bronze medal. Ranked 465th out of 6,125 teams in the ROGII Wellbore Geology Prediction competition.
___
74 / 2963
Kaggle — 2026 NeuroGolf Championship
Silver medal. Ranked 74th out of 2,963 teams in the 2026 NeuroGolf Championship.
___
244 / 3357
Kaggle — Santa 2025
Bronze medal. Ranked 244th out of 3,357 teams in the Santa 2025 competition on Kaggle.
___
1st
Design Competition — Taylor's University
First place in the design competition conducted at Taylor's University, Malaysia. MADE program 1st edition.
___
29 / 4206
Kaggle — Diabetes Prediction 2025
Placed 29th out of 4,206 teams on the Kaggle Diabetes Prediction Challenge, 2025 edition.
___
Languages
Python, C++, SQL, R
Tools
Linux, Git, Docker, GitHub, PowerBI, AWS, Supabase, Vercel, Modal
Libraries
Pandas, NumPy, Matplotlib, PyTorch, OpenCV, Scikit-learn, Seaborn, HuggingFace, vLLM, Unsloth
Focus
Optimization, deployment, real-time inference, data analysis
Mail GitHub Kaggle