Resume / Updated November 2026
Aditya Anil
Machine Learning Engineer
Machine Learning Engineer experienced in building and deploying production AI systems, APIs, and computer-vision pipelines. At IHX, re-engineered a production product supporting approximately 90% of company revenue, achieving 99% table-classification precision and reducing processing time by 50%; in a personal project, built multimodal LangGraph agents with validation, tracing, and cost-aware model selection. Skilled in Python, PyTorch, AWS, FastAPI, and model evaluation, with ongoing research in interpretability at UTS.
Experience
Research Assistant
University of Technology Sydney (UTS) / Sydney, Australia
- Investigate mechanistic interpretability of EEG foundation models, studying how pretrained EEG transformers (CBraMod) encode neurological states with Cross-Layer Transcoders, sparse features, and activation-level analysis of drowsiness-related features.
- Built end-to-end EEG machine-learning pipelines for preprocessing, model adaptation, feature extraction, and evaluation; benchmarked linear probing, partial fine-tuning, LoRA, and full fine-tuning on SEED-VIG.
- Evaluated 153,600-dimensional representations across 12 transformer layers using hypothesis tests, effect sizes, permutation testing, and subject-independent validation.
- Co-authored AE-UVNet, an attention-enhanced hybrid U-Net/V-Net architecture for multimodal brain tumour segmentation, published in Springer Lecture Notes in Computer Science (2026).
- Investigated diffusion-based object detection for complex aerial imagery, focusing on low-resolution objects, occlusions, and challenging visual conditions.
- Co-authored three research publications spanning medical imaging, deep-learning model generation using matrix factorisation, and AI applications in clinical healthcare.
Academic Tutor - 32130 Fundamentals of Data Analysis
University of Technology Sydney (UTS) / Sydney, Australia
- Evaluated assignments and led oral defenses to assess students' understanding of data analysis.
- Applied probability, statistics, and data interpretation while maintaining academic integrity.
- Supported tutorials by explaining statistical concepts and providing actionable feedback to strengthen analytical reasoning.
Machine Learning Engineer
IHX / Bangalore, India
- Re-engineered a production document-intelligence product supporting approximately 90% of company revenue, improving processing efficiency, scalability, and operational reliability.
- Trained, fine-tuned, and integrated Donut, LayoutLMv3, Table Transformer (TATR), and DETR for healthcare document classification and structured information extraction.
- Built and optimized batch and real-time inference pipelines across CPU and GPU environments, balancing latency, throughput, and compute costs.
- Achieved 99% precision in table classification and reduced pipeline execution time by 50% through parallel processing and workflow optimization.
- Built Python and FastAPI REST APIs to serve AI-powered document processing and integrate it with internal products and engineering teams.
- Supported production releases on AWS and GCP, administered servers via SSH, and monitored production systems with Grafana.
- Developed scraping and data-processing pipelines across four pharmacy platforms, creating a structured pharmaceutical data store and a fraud-detection model.
- Mentored and onboarded an ML engineer, documented production workflows, and transferred knowledge for critical systems.
Open Source Contributor
docTR / Remote
- Contributed to docTR, a deep-learning OCR framework, improving text detection and recognition accuracy, model efficiency, and document understanding capabilities.
- Supported optimization of computer-vision and deep-learning pipelines for robust, high-performance optical character recognition.
Selected projects
PitchSense - AI Football Analytics & Tactical Tracking
Built an end-to-end computer-vision system that converts standard football match footage into tactical telemetry without GPS trackers or specialist hardware. Developed player and ball tracking, pitch registration, and an analytics interface; won Best Innovation/Research Project at the 2026 UTS AI Showcase.
Multimodal Memory & Agentic Content System
Built Pixel Sprites, a Python multimodal AI application using LangChain and LangGraph, with specialized agents for task classification, media understanding, content planning, keyword generation, and output validation across images, video, and audio. Added structured validation retries, cost-aware model tiers, tracing, media indexing, and Supabase persistence.
AI Car Insurance Advisor
Built an end-to-end insurance recommendation system combining driver-risk prediction, premium estimation, policy ranking, and evidence-backed recommendations. Integrated Qdrant retrieval over Product Disclosure Statements and SQLite interaction tracking.
Habit Tracker & AI-Powered Task Categorizer
Built and currently run an LLM-powered categorization agent for Google Tasks, a modular Gemini/OpenAI provider architecture, a Flutter productivity app, and daily categorization automation with GitHub Actions.
Education
Feb 2025 - Oct 2026
University of Technology Sydney (UTS)
Master of Artificial Intelligence
84.33Jun 2019 - Jun 2023
Amrita Vishwa Vidyapeetham
Bachelor of Technology, Computer Science (Artificial Intelligence)
87.00Skills
Agentic AI & LLM Engineering
LangGraph / LangChain / Multi-agent systems / Stateful agent orchestration / Conditional routing / Agent memory / LLM tool integration / Structured outputs / Pydantic / Validation and retries / OpenRouter / OpenAI API / Gemini API
AI Observability, Evaluation & Cost Optimization
LangSmith tracing / LLM workflow debugging / Model evaluation / Output validation / Cost-aware model routing / Configurable LLM tiers / Grafana / Latency optimization / CPU/GPU workload scheduling
Production AI Engineering & MLOps
Python / FastAPI / REST APIs / Deep-learning deployment / Batch and real-time inference / Production monitoring / Git / GitHub Actions / CI/CD / Linux / AWS / GCP
Deep Learning & Multimodal AI
PyTorch / TensorFlow / Hugging Face Transformers / Fine-tuning / LoRA / Vision-language models / Donut / LayoutLMv3 / DETR / Table Transformer / YOLO / OpenCV / Whisper / InsightFace / OCR
RAG, Data & Persistence
Retrieval-Augmented Generation / Qdrant / Supabase / Firebase / SQLite / SQL / NumPy / pandas / SciPy / Vector embeddings / Metadata extraction / Structured JSON
Applications & Research
Streamlit / Flutter / scikit-learn / Statistical analysis / Mechanistic interpretability / Sparse feature analysis / Model benchmarking