Aashish Dhawan

AI Researcher

Gainesville, Florida, United States4 yrs 4 mos experience

AI ML PractitionerAI Enabled

Key Highlights

Expert in Machine Learning and Multimodal AI.
Proven track record in enhancing indigenous language translation.
Experience in developing scalable AI solutions.

Stackforce AI infers this person is a Machine Learning and AI specialist with a focus on Multimodal applications.

Contact

Skills

Core Skills

Machine LearningNatural Language Processing (nlp)Artificial Intelligence (ai)Machine Learning AlgorithmsPattern Recognition

Other Skills

Predictive ModelingPython (Programming Language)Data MiningApplied MathematicsAssistant TeachingComputer VisionCRFsMySQLLangChainOpenAI APIFAISSembedding modelRetrieval-Augmented Generation (RAG)Large Language Models (LLM)MLOps

About

PhD student in Computer Science at the University of Florida working at the intersection of Machine Learning, Multimodal AI, Vision, and NLP. I’ve built and deployed real systems across low-resource machine translation, computer vision, generative AI, and LLM applications — from improving indigenous language translation using multimodal learning to shipping production AI features for real users. Previously: - Research Intern at ISRO (satellite image processing) - Visiting Researcher at the University of Sydney (domain adaptation under Prof. Dacheng Tao) What I care about: Turning research into usable systemsMaking AI practical, explainable, and scalable Tech I work with: Tensorflow, PyTorch, LLMs, RAG, Diffusion, GANs, CV, NLP. Always open to: research collaboration, startup work, and high-impact ML roles.

Experience

4 yrs 4 mos

Total Experience

4 yrs 4 mos

Average Tenure

4 yrs 4 mos

Current Experience

University of florida

2 roles

Graduate Research Assistant

Aug 2025 – Present · 9 mos

Machine Learning Researcher

May 2024 – Aug 2025 · 1 yr 3 mos

Advisor - Dr. Daisy Wang
Enhanced machine translation for indigenous low-resource (endangered) languages by fine-tuning LLMs like Meta's mBART, a multilingual sequence-to-sequence denoising auto-encoder, achieving a 10-15% increase in BLEU scores.
Generated aligned text-image datasets utilizing diffusion models and GANs, enhancing the training corpus and enabling robust multimodal machine translation for underrepresented languages.
Integrated visual features by aligning mCLIP embeddings with textual data, resulting in an additional 10% improvement in translation accuracy, demonstrating the efficacy of multimodal approaches in low-resource settings.
Applied object-level visual context modeling, implementing selective masking of irrelevant visual features to improve translation grounding and precision, thereby enhancing the contextual relevance of translations.
Conducted comprehensive studies to evaluate the impact of multimodal integration and object-level masking, providing insights into effective strategies for improving machine translation in low-resource, indigenous language contexts.

Python (Programming Language)Natural Language Processing (NLP)

Atmosphere apps

AI Solutions Consultant

Oct 2023 – Apr 2024 · 6 mos · Gainesville, FL · Remote

Built a GPT-powered drug recommendation system, improving prescription relevance by ~20%.
Developed a multilingual video transcription + translation pipeline using Whisper + GPT-4 with 95% accuracy across 10+ languages.
Performed prompt engineering to evaluate and optimize existing prompts, leading to 30% better contextual relevance and 20% reduction in response errors in model outputs.
Handled model selection, evaluation, and optimization, leveraging AWS infrastructure for scalable and high-performance AI solutions.

Data MiningArtificial Intelligence (AI)

University of florida

3 roles

Adjunct Instructor

Aug 2023 – Present · 2 yrs 9 mos

Instructor for Programming Languages (COP 4020) and Intro to Machine Learning (CIS 4930).
Design full course pipelines: syllabus, projects, exams, and evaluation.
Lead and mentored 5+ TAs and peer mentors.
Supervised real-world student projects in:
Chatbots
Stock sentiment analysis
Fraud detection

Machine LearningPredictive Modeling

Graduate Teaching Assistant

Jan 2021 – May 2023 · 2 yrs 4 mos · United States

1. Operating Systems COP 4600 at UF (Spring 2021) - Lead a team of 9, Discussions, Office hours and grading.
2. Programming Language Concepts COP 4020 (Summer 2021) - Office hours and grading.
3. Mobile Computing CNT 5517 (Fall 2021) - Taught Android to Graduate students including other TA duties
4. Programming Language Concepts COP 4020 (Spring 2022-May 2023), Lead the TA teams, grading, taking office hours and other essential duties

Applied MathematicsAssistant Teaching

Graduate Research Assistant

Jan 2021 – May 2023 · 2 yrs 4 mos · United States

Developed and implemented a PROGAN framework to generate multispectral images, encompassing satellite and medical imagery, significantly enhancing data availability for cancer research initiatives
Spearheaded research efforts focused on leveraging Neural Networks for Quantum mechanics tasks, notably reducing computation time by 82 percent in approximating elemental patterns through Schroedinger’s equation calculations
Led Mitsuba2 compilation for fluorescence rendering and predictive modeling for material scattering coefficients.Validated through 3D printed object experiments, reducing evaluation time by 70%

Machine Learning AlgorithmsComputer Vision

University of sydney

Visiting Researcher

Dec 2019 – Mar 2020 · 3 mos · Greater Sydney Area

Advisor - Dr. Dacheng Tao
Built multi-source domain adaptation models across 6 domains, 345 classes, 600k samples.
Used moment matching + adversarial learning (GRL) for domain-invariant features.
Achieved 71–98% accuracy across the domains.

Machine Learning AlgorithmsComputer Vision

Space applications centre, isro

Research Intern

Jun 2018 – Aug 2018 · 2 mos · Ahemdabad

Advisor - Dr. SP Vyas
Engineered a satellite image segmentation pipeline by integrating CNN-based segmentation with CRF-based post-processing, improving segmentation accuracy by 30% on space imagery from the LISS-3 sensor on IRS-2 (ResourceSat-2) and the POTSDAM aerial photography dataset.
Optimized computational efficiency, reducing image processing time from 36 hours to 0.2 seconds by redesigning key components and employing efficient graphical models for post-processing.
Developed custom post-processing algorithms using CRFs to enhance boundary delineation in high-resolution satellite imagery, ensuring improved clarity for downstream applications in remote sensing.
Published research findings at INDIACom-2019, detailing novel techniques for satellite image segmentation and post-processing, contributing to advancements in the field of remote sensing.
Utilized advanced tools and libraries such as PyTorch, NumPy, and ArcGIS to build scalable image processing models, demonstrating proficiency in graphical models, pattern recognition, and satellite data analysis.

CRFsPattern Recognition