Senior ML Engineer

This is a fully remote position, open to applicants in Serbia, +1 more country.

📋 Description

• Establish the machine learning strategy by selecting strategic initiatives and aligning each with specific business requirements.

• Oversee the complete process of large language model (LLM) post-training and model adaptation, from constructing datasets to fine-tuning and optimizing preferences.

• Create and develop agent harnesses and systems that utilize tools with LLMs.

• Enhance multi-agent development workflows to boost team productivity.

• Construct evaluation frameworks and LLM-as-a-judge systems that forecast real-world performance.

• Elevate dialogue quality by focusing on memory, emotional intelligence, persona consistency, safety, and preventing looping behavior.

• Deliver production systems that significantly influence user experience, retention, and monetization.

⛳️ Requirements

• Expertise in Python: producing clean code, implementing typing, async programming, multiprocessing, profiling, debugging, and handling both notebooks and production code.

• Proficiency with Transformers.

• Experience in training and fine-tuning LLMs.

• Knowledge of SFT, LoRA/QLoRA, and PEFT.

• Familiarity with preference optimization techniques: DPO / ORPO / SimPO / CPO / GRPO or a commitment to rapidly gain expertise.

• Skills in building datasets for LLM training, including cleaning, filtering, deduplication, preference pairs, synthetic data, and hard negative mining.

• Experience in LLM evaluation methods: offline evaluations, pairwise evaluations, LLM judges, and human annotation.

• Capabilities in agent harness design: managing tool calls, function calls, structured outputs, retries, routing, state management, memory/context, and guardrails.

• Experience with production LLM integration involving APIs, streaming, batching, fallbacks, and managing cost/latency trade-offs.

• Understanding of observability for agents, including traces, tool-call logs, transcript analysis, and failure clustering.

• Strong grasp of transformer architecture and contemporary LLM behaviors.

• Precision with data types, devices, and batching: fp32/fp16/bf16, quantization, CUDA memory management, and preventing dtype mismatch errors.

• Proficiency with Git, Docker, Linux, and basic MLOps.

• Fluency in Russian.

• Availability to work within CET (±2) hours.

• Preferred: experience beyond text, such as in computer vision, image generation, multimodal applications, maintaining long-running conversations and character consistency, familiarity with vLLM / TGI / SGLang, DeepSpeed / FSDP / Accelerate, quantization, and safety classifiers.

🏝️ Benefits

• REMOTE OPPORTUNITY for full-time work.

• 28 calendar days of vacation annually.

• 7 wellness days off each year.

• Bonuses up to $5000 for referring successful candidates for positions within the company.

• 50% reimbursement for professional training, international conferences, and meetings.

• Corporate discounts available for English language lessons.

• Health benefits including reimbursement of up to $1,000 gross per year for personal health insurance or doctor’s fees for yourself and close relatives if corporate medical insurance is not applicable.

• Provision of a well-equipped workplace and necessary equipment in offices or co-working spaces.

• Reimbursement of workplace expenses up to $1000 gross once every 3 years for co-working rentals or home office equipment.

• An internal gamified gratitude system that allows for bonuses redeemable for merchandise, team-building activities, massage certificates, and more.