experience

here I have highlighted some of my relevant experience.

🧑🏻‍💻

March 2025 - Present

Madrid, Spain - onsite

Conducting research under Dr. Luis Fernando D'Haro within the EU Horizon INNOVATRAD project — developing multi-agent microservices pipelines for Spanish-English simultaneous interpretation with domain adaptation on Google Cloud Run
Designed a training-free terminology adaptation pipeline for interpreted speech translation, discovering domain terms from ASR uncertainty and biasing decoding to close the terminology gap against professional interpretation, with no model fine-tuning (submitted to ICASSP 2027)
Delivered a live multimodal museum visitor assistant fusing speech, vision and text via a context-aware LLM with a Neo4j knowledge-graph RAG backend
Contributed to EU Horizon funded ASSIST (101201944) and ASTOUND (101071191) projects, focused on enhancing museum visitor experience through multimodal, personalized, and context-aware interactions
Designed and deployed an end-to-end agentic framework using microservices; developed personalized chatbot architectures and managed containerized execution on Google Cloud Run
Presented the System Demo at the EU Horizon Transition Project Final Review of ASTOUND, representing the lab's contributions before the evaluation panel
Two granted software IP registrations with the Spanish patent office: "Multimodal-AIC", for automatic analysis of attention, intentionality and contextualization from multimodal signals, and "ART_MIRA", covering the multimodal museum visitor assistant above

🧑🏻‍💻

July 2024 - March 2025

Singapore - onsite

Collaborated with Dr. Eng Siong Chng and team on cross-domain synthetic speech detection, designing and benchmarking four novel architectures across multiple spoofing corpora
Architectures developed: Internal Ensemble Learning, Progressive Enhancement Learning, Group Propagation, and Domain Discriminator

🧑🏻‍💻

August 2024 - February 2025

Kathmandu, Nepal - onsite

Conducted research under Dr. Bishesh Khanal on developing comprehensive parallel corpora for Nepali language NLP applications
Served as Teaching Assistant for the prestigious Annual Nepal AI School, supporting courses on Geometric Deep Learning led by Dr. Michael Bronstein and LLMs and Mixture of Experts led by Dr. Chetan Arora
Led the "nepSWOR" project in collaboration with NAAMII's startup Diyo, building models, datasets, and evaluation frameworks for Nepali ASR and TTS systems across 9 regional dialects and accents

🧑🏻‍💻

Jan 2024 - July 2024

Singapore - onsite

Created and curated a diverse synthetic audio dataset using various voice cloning technologies for training and evaluation of deepfake detection systems
Fine-tuned WavLM with LoRA for Speech Emotion Recognition (SER), cutting trainable parameters by over 90% without accuracy loss, and implemented self-supervised learning models (WavLM, Wav2vec2) with RawBoost augmentation for robust spoofing detection
Achieved Rank 3 in the SVDD 2024 Challenge by developing an ensemble system using speech foundation models with a novel Squeeze-and-Excitation Aggregation (SEA) method, reaching a 1.79% pooled equal error rate for singing voice deepfake detection GitHub

🧑🏻‍💻

May 2023 - June 2023

Kathmandu, Nepal - onsite

Developed a network monitoring system using Python's Django, PostgreSQL, Redis, and Docker for enterprise-level device management
Integrated DevOps tools (Zabbix, Nornir) with custom Django UI to restructure internal ticketing system, reducing manual email-based workflows by one-third and improving operational efficiency

🧑🏻‍💻

April 2021 - June 2021

Kathmandu, Nepal - onsite

Conducted research in Mathematics and Cryptography, developing curriculum for the 'Cryptography for Everyone' course
Successfully launched and taught the course with 70+ student enrollments, covering asymmetric/symmetric key cryptography, hashing, key exchange, and network vulnerabilities