Full Stack Developer

We're seeking an experienced LLM/ML Specialist with deep expertise in LLaMA models or other open source and Retrieval-Augmented Generation (RAG) systems. The ideal candidate will have strong skills in model fine-tuning, prompt engineering, and production deployment of language models. You'll build and optimize RAG pipelines, implement vector databases, and develop efficient inference solutions. Requirements include 2+ years of LLM experience, Python proficiency with PyTorch/Hugging Face, and demonstrated projects involving LLaMA models.   Technical Skills: Deep understanding of LLaMA model architecture and its variants Experience fine-tuning and adapting LLaMA models for specific applications Proficiency in Python and ML frameworks (especially PyTorch and Hugging Face) Knowledge of prompt engineering specific to LLaMA models Experience with efficient inference and quantization techniques for LLaMA Understanding of model deployment and optimization for large language models Expertise in Retrieval-Augmented Generation (RAG) systems and architectures Experience implementing vector databases and similarity search techniques Knowledge of document chunking and embedding strategies for RAG Familiarity with evaluation metrics for RAG systems Qualifications: Machine Learning, NLP, Computer Science, or related field 2+ years of experience working with large language models, preferably LLaMA Strong mathematical background in statistics and probability Demonstrated projects involving LLaMA model adaptation or deployment Excellent communication skills to explain complex concepts Proven experience building and optimizing RAG pipelines in production environments Preferred: Experience with PEFT methods (LoRA, QLoRA, etc.) for LLaMA models Experience with fine-tuning Understanding of model limitations and ethical considerations Experience with LLaMA integration into production systems Familiarity with open-source LLM ecosystems Experience with hybrid search methodologies (dense + sparse retrieval) Knowledge of context window optimization techniques for RAG systems Experience with multi-stage retrieval architectures Responsibilities: Design and implement LLaMA-based solutions for real-world applications Develop, optimize, and deploy RAG systems using vector databases and embedding strategies Fine-tune LLaMA models using techniques like LoRA and QLoRA Create efficient prompt engineering strategies for specialized use cases Implement and optimize inference pipelines for production environments Design evaluation frameworks to measure model and RAG system performance Collaborate with engineering teams to integrate LLMs into product infrastructure Research and implement the latest advancements in LLM and RAG technologies Document technical approaches, model architectures, and system designs Mentor junior team members on LLM and NLP best practices What We Offer: Opportunity to work on cutting-edge LLM and RAG applications Access to computational resources for model training and experimentation Collaborative environment with other ML/AI specialists Flexible work arrangements with remote options Competitive salary and comprehensive benefits package Professional development budget for conferences and courses Clear career progression path for AI specialists Chance to contribute to open-source LLM ecosystem Balanced workload with dedicated research time Inclusive culture that values diverse perspectives and innovative thinking

What you will do: Build and ship core ML models for a speech-driven behavioral engine. Own end-to-end modeling from raw, long-form audio and layered annotations to production inference. Design audio features/embeddings, train and evaluate a suite of models, and deliver reproducible pipelines that meet accuracy, robustness, latency, and cost targets. Your skill set and experience: 5+ years building production ML systems, including 2+ years in speech/audio. Speech & signal processing: VAD, diarization, segmentation, denoising, spectral features (log-mel/MFCC), prosody (pitch/energy), long-form audio handling. SOTA audio models & embeddings: Wav2Vec2, HuBERT, wavLM (or similar); fine-tuning/self-supervised learning; contrastive/metric learning for downstream tasks. Data excellence: SQL, Python data stack (Pandas/Polars), ETL for audio+metadata, stratified sampling, leakage prevention, feature stores. ML Training: PyTorch, Hugging Face Transformers/Hub, mixed precision, hyperparameter tuning, transfer learning, cross-validation. Evaluation discipline: golden sets, robust speaker/content splits, ROC/PR/calibration, fairness/bias checks, ablations, drift/shift detection on embeddings and audio quality. MLOps, serving & reproducibility: FastAPI/gRPC around HF/torchaudio models, experiment tracking (W&B/MLflow), artifact/model versioning, CI/CD, observability, scalable batch/streaming inference. Proven ability to create and document novel IP (methods, architectures, or training/eval techniques) with clear prior-art awareness. Nice to have: Tooling: SpeechBrain, Lightning, OpenSMILE/Praat, Kaldi/Conformer/Emformer, Label Studio. Multimodal: ASR (e.g., Whisper) + paralinguistic features; emotion/prosody modeling; speaker embeddings (x-vectors, ECAPA-TDNN). Performance & deployment: quantization/distillation, Triton/CUDA basics, distributed training, real-time/streaming inference, on-device DSP (Rust/C++). Publications/patents/competition results demonstrating novel audio modeling work.

Full-time (40h/week), Remote  About Us We are a technology company on a mission to transform how organizations use data. We believe that for businesses to lead their industries, they must effectively harness their data. Our innovative, AI-powered analytics platform was built to make data-driven insights universally accessible, eliminating the need for complex infrastructure or large specialist teams. We empower non-technical users to connect to hundreds of data sources, ask questions in plain English, and get immediate, actionable insights. By unifying disparate data and enabling easy collaboration, we help businesses make faster, smarter decisions and turn their data into a pivotal competitive advantage. About the role We are seeking a Customer Success Engineer to handle customer-specific technical implementations and platform work. This is a crucial, forward-facing technical role designed to support our customers by addressing their unique account issues and technical needs, complementing our existing data team. This position will report directly to the CTO and work closely with internal stakeholders like the Product Manager and the Customer Success team. The primary focus will be on technical execution, with the potential for limited direct client communication via email or tickets. You will be instrumental in ensuring our customers are successfully onboarded and supported, directly impacting their success and our company's growth. Key responsibilities: Configure, debug, and manage ETL pipelines for new customer implementations. Build and maintain custom Airbyte connectors (Python or low-code/no-code) to meet specific customer data integration needs. Troubleshoot and resolve issues with Fivetran connectors. Provide AWS infrastructure support for customer environments, particularly managing S3 and IAM roles for secure data transfer Handle API integration work for customer accounts. Collaborate with and support our existing data team contractors. Utilize AI tools to debug SQL queries and troubleshoot various technical issues. Perform data quality assurance and validation to ensure data integrity. Set up customized BI dashboards for customer accounts within our platform. Review user LLM session logs to diagnose issues with our RAG pipeline and create detailed tickets for the engineering team. Requirements: Mid-level or higher engineering experience with a strong background in data engineering and ETL processes. Proficiency with Fivetran, Airbyte or equivalent ETL tools. Proficiency with AWS, specifically S3 and IAM roles. Solid experience with API integration, and debugging. Basic to advanced SQL proficiency for debugging and data validation. Proven ability to work independently and take ownership of customer-facing technical issues. Comfortable debugging issues throughout a data platform stack (from front-end to database and infrastructure) and writing clear, actionable bug tickets for the development team. Tools: Cloud: AWS (S3, IAM) Data Tools: Fivetran, Redshift, pgVector/Postgres, Metabase Languages: Ruby, Python Frontend: React (experience is helpful) Authentication: Keycloak AI: OpenAI, Anthropic Project Management: Trello Nice to have: Familiarity with AI/LLM tools and debugging (or enthusiasm to learn) Existing Airbyte/Fivetran/general ETL connector development experience Familiarity with data transformation tools and data modeling (eg. dbt) More advanced SQL and data analysis skills We would want at least 4 hours of overlap in the morning (9AM-1PM) as we have people in other timezones where the morning makes sense. Ideally would be US hours overlap (9AM-6PM or thereabouts) as reactive issues can come in from clients throughout the day.