AI & Data
NLP Challenges
NLP challenges put you inside the work of teaching machines to read and make sense of language. You'll develop skills in Natural Language Processing fundamentals, Text Tokenization and Word Embeddings, and tasks like Named Entity Recognition and Sequence labeling using NLTK.
From there you'll handle the harder edges — Encoder fine-tuning (BERT family) with Hugging Face Transformers, Custom tokenization, Relation extraction, Information Retrieval, and Multilingual NLP — building Knowledge Representation the way real NLP teams do. Each challenge you solve earns a verified credential you can share with recruiters.
Recommended Challenges
· Hugging Face Transformers Clear- CodeIntermediateNew
LoRA Fine-Tune a 7B LLM for Legal-Clause Extraction
You receive a curated extraction dataset (2,000 train, 500 val, 500 test contracts with span-level labels across 12 clause types) and a fine-tunable 7B base model (e.g., Llama-3…
- Fine Tuning
- Fine Tuning
- Parameter Efficient Tuning
Fine-Tuning Large Language Models - CodeIntermediateNew
Build a Multimodal Generation Pipeline for a Tourism Operator
You receive 40 sample 30-second videos shot by tour guides, the operator's brand voice doc, and SEO keyword lists for EN/PT/ES. Build a pipeline that (1) extracts a representati…
- Multimodal Generation
- Vision Language Models
- LLM Inference
Generative AI - ResearchSeniorNew
Compare RNN vs Transformer for Long-Sequence Modeling
Pick a public trajectory dataset (e.g., Argoverse 2, Waymo Open, or ETH-UCY). Implement three models with comparable parameter counts (around 5M each): an LSTM baseline, a vanil…
- Hugging Face Transformers
- Rnn
- State Space Models
Neural Networks for NLP - ResearchBeginnerNew
Drug-Repurposing Candidate Screen with Embedding Similarity
You receive (1) a list of 15 known therapeutic candidates (SMILES + ChEMBL identifiers) for a single rare disease, (2) a database of about 4,500 marketed drugs (SMILES + ATC cod…
- Molecular Embeddings
- Similarity Search
- Transfer Learning
Machine Learning for Healthcare and Biomedicine Practice your coursework on real scenarios.
Every challenge is shaped from real industry context — not generic exercises. The work mirrors what your degree prepares you for.
Why Ewance
- ResearchIntermediateNew
Probe a Pretrained Encoder for Linguistic Knowledge
Take BERT-base (or DeBERTa-v3-base). Run layer-wise probes across at least 3 linguistic tasks: part-of-speech tagging, dependency arc classification, and semantic role labeling.…
- Interpretability
- Probing
- Hugging Face Transformers
Neural Networks for NLP - CodeIntermediateNew
Build a Sequence Model for Sign-Language Word Recognition
You receive about 12,000 short (1-3s) webcam clips covering a 50-word vocabulary, with body+hand pose features pre-extracted (e.g., MediaPipe Holistic landmarks per frame). Buil…
- Sequence Models
- Hugging Face Transformers
- Pose Estimation
Machine Perception - CodeIntermediateNew
Build a Small Transformer from Scratch and Train It on Code
Implement multi-head self-attention, RMSNorm, rotary positional embeddings, and a causal LM head from scratch — no Hugging Face shortcuts for the model code (you may use Hugging…
- Hugging Face Transformers
- Self Attention
- Pytorch Or Tensorflow
Neural Networks for NLP - CodeIntermediateNew
Adapt Machine Translation to a Niche Domain
Pick an open MT base (NLLB-200 or a strong open M2M model). Build a parallel corpus of around 8,000 sentence pairs from the company's bilingual safety standards. Fine-tune on th…
- Machine Translation
- Domain Adaptation
- Hugging Face Transformers
Natural Language Processing - Browse challenges
Explore role
Product Manager
Ship product that solves real user problems. Combine user research, prototyping, and stakeholder alignment to turn ambiguous briefs into measurable wins — the role at the centre of modern software teams.
- CodeIntermediateNew
Train a Domain-Specific Reranker for a Legal-Tech Search Box
You receive 20,000 (query, document, relevance-label) triples from the firm's contract corpus. Fine-tune a small cross-encoder (e.g., ms-marco-MiniLM-L-6-v2 or BAAI/bge-reranker…
- Cross Encoder Reranker
- Fine Tuning
- Ir Evaluation
Information Retrieval and Search - DesignBeginnerNew
Build an Attention-Visualization Tool for Translation Quality Audit
You will load a small open-source EN-FR transformer (e.g., Helsinki-NLP Opus-MT-en-fr), build a Streamlit or Gradio demo that lets the user paste English source, see the French …
- Attention Mechanisms
- Neural Mt
- Tool Design
Machine Translation - ResearchSeniorNew
Plan a Parameter-Efficient Fine-Tuning Strategy for a Big-Tech AI Lab
You will produce (1) a 6-page survey of four PEFT methods (LoRA, adapters, prefix tuning, IA3) with their strengths, weaknesses, and parameter footprints, (2) a one-page decisio…
- Parameter Efficient Fine Tuning
- Transfer Learning
- Fine Tuning
Meta-Learning, Transfer Learning, and Multi-Task Learning - CodeIntermediateNew
Build a Domain-Specific Named-Entity Recognizer for Legal Contracts
Start from a strong English NER base (spaCy transformer or LegalBERT). Fine-tune on a provided 1,200-contract labeled dataset for the 9 entity types. Handle long contracts (ofte…
- Named Entity Recognition
- Sequence Labeling
- Domain Adaptation
Natural Language Processing Build a verifiable portfolio.
Submissions become evidence. Reviewers with shipping experience score against a rubric; the result becomes a credential anyone can verify.
Why Ewance
- CodeIntermediateNew
Instruction-Tune a Small Model for an Edtech Tutor
You receive a 1.5B base model (e.g., SmolLM-1.7B or Qwen-1.8B), permission to use 2 hours of a rented A100, and a curated seed of around 5,000 math-tutoring dialogues. Augment w…
- Instruction Tuning
- Fine Tuning
- Dataset Curation
Fine-Tuning Large Language Models - CodeIntermediateNew
Fine-Tune a Sequence-to-Sequence Model for Code-Doc Generation
Take a small base model (CodeT5+ or a distilled CodeLlama-Instruct). Build the dataset by mining around 8,000 high-quality function-docstring pairs from permissively-licensed Py…
- Seq2seq
- Hugging Face Transformers
- Fine Tuning
Neural Networks for NLP - CodeIntermediateNew
Fine-Tune a Transformer for Customer-Support Triage at an Enterprise AI Vendor
You receive 240,000 labeled support tickets across 14 queues, with English, Bahasa Indonesia, and Tagalog. Fine-tune a multilingual transformer encoder (XLM-RoBERTa-base is a st…
- Hugging Face Transformers
- Fine Tuning
- Multilingual NLP
Deep Learning - CodeIntermediateNew
Domain-Adapt an NLP Pipeline from News to Customer-Support Tickets
You receive 30,000 anonymized customer-support tickets (PT-BR + ES) plus the news-trained NER and intent models. Apply continued pretraining of a multilingual encoder (e.g., XLM…
- Transfer Learning
- Domain Adaptation
- Continued Pretraining
Meta-Learning, Transfer Learning, and Multi-Task Learning - CodeIntermediateNew
Build a 30-Day Readmission Risk Model on De-Identified EHR Data
You receive a curated MIMIC-style de-identified EHR cohort (about 28,000 admissions, demographics, comorbidities, labs, prior-admission counts) with 30-day readmission labels. T…
- Ehr Modeling
- Risk Stratification
- Model Calibration
Machine Learning for Healthcare and Biomedicine - CodeBeginnerNew
Train a Word-Alignment Model for Low-Resource Catalan-Aranese
You receive a 35,000-sentence Catalan-Aranese parallel corpus plus a 1,200-pair manually annotated word-alignment test set. Train (1) a classic statistical alignment baseline (e…
- Alignment
- Neural Mt
- Low Resource Mt
Machine Translation - CodeIntermediateNew
Build a Multilingual Customer-Email Classifier
You receive 28,000 labeled emails (skewed toward English and Mandarin). Try at least two approaches: (1) a fine-tuned multilingual transformer (XLM-RoBERTa or mDeBERTa) and (2) …
- Text Classification
- Multilingual NLP
- Hugging Face Transformers
Natural Language Processing - CodeIntermediateNew
Fine-Tune a Small Transformer for Legal-Domain EN-DE Translation
You receive a 120,000-segment parallel EN-DE legal corpus and a held-out 1,000-segment test set with reference translations. Fine-tune a small pretrained Transformer (e.g., NLLB…
- Neural Mt
- Hugging Face Transformers
- Fine Tuning
Machine Translation - ResearchIntermediateNew
Multi-Task Learning for a Healthtech Triage Model
You receive 40,000 anonymized de-identified intake-form records with two labels: urgency tier (4 classes) and routed sub-specialty (12 classes). Train (1) two independent classi…
- Multi Task Learning
- Transfer Learning
- Hugging Face Transformers
Meta-Learning, Transfer Learning, and Multi-Task Learning - ResearchIntermediateNew
QLoRA Fine-Tune for a Customer-Support Domain Assistant
You receive 8,000 anonymized support ticket pairs (question -> agent response), the company's product documentation (around 600 pages), and a strong RAG baseline already running…
- Qlora
- Fine Tuning
- RAG Architectures
Fine-Tuning Large Language Models - CodeIntermediateNew
Train a Sequence Model for Wearable-Telemetry Sleep Staging at a Healthtech
You receive 220 nights of wearable telemetry from 60 subjects with PSG ground-truth labels. Train three sequence models: an LSTM baseline, a 1D-CNN+GRU hybrid, and a small trans…
- Sequence Models
- Lstm
- Hugging Face Transformers
Deep Learning - AnalysisIntermediateNew
Catastrophic-Forgetting Audit on a Domain Fine-Tune
You receive the fine-tuned 7B chemistry model and its base, plus a benchmark basket (MMLU subset, GSM8K, IFEval, a small instruction-following set). Run all 4 benchmarks on both…
- Catastrophic Forgetting
- LLM Evaluation
- Fine Tuning
Fine-Tuning Large Language Models
How it works
From brief to credential, in six steps.
Step 01
Browse challenges aligned to your studies.
Step 02
Accept the one that fits your goals.
Step 03
Work through it with AI Copilot guidance.
Step 04
Submit for structured evaluation.
Step 05
Earn a verified credential.
Step 06
Add it to LinkedIn with one click.
Industry teams behind a decade of practitioner briefs
Hiring from this pool?
Sponsor a challenge and meet candidates through actual work.
Industry teams can shape briefs around the skills they hire for, then evaluate students on rubric-scored deliverables — not resumes.