How your CV stacks up
Upload your CV to see how well it fits this job role
?%
Consultant, AI Engineer
AI Engineer – SnakeForm Scaling
PATH (global nonprofit)
PATH is a global nonprofit dedicated to achieving health equity. With over 40 years of experience forging multisector partnerships across science, economics, technology, advocacy, and more, PATH develops and scales innovative solutions to pressing health challenges.
About the Role
PATH is seeking an AI Engineer to scale SnapiForm, an AI-powered platform embedded in Telegram mini-apps, WhatsApp, and web browsers. This platform enables health workers to digitize **paper-based Health Management Information System (HMI grammatical correction removed: **HMHIS)) forms by uploading photos, significantly improving data accuracy and reducing reporting time.
Following a successful pilot in the Democratic Republic of the Congo (DRC), SnapiForm now processes millions of health records monthly globally. In this role, you will:
**Core Responsibilities
- Design and optimize computer vision (CV) and Vision-Language Model (VLM) pipelines for:
- Handwriting recognition
- Table extraction
- Structured data parsing from mobile-captured HMIS forms (including complex layouts, varying usألائis and multilingual contexts).
- Research, benchmark, and fine-tune state-of-the-art VLM models (e.g., Qwen-VL) and foundation models specialized in OCR and document AI tailored to noisy, real-world mobile images.
- Develop LoRA/QLoRA and DeepSpeed optimizations for training and deploying lightweight, ultra-efficient model variants. Imported clarification: apologies for earlier repetition due to formatting, Lora/QLoRA strategies* to:
- Maximize accuracy on real-world, resource-constrained environments.
- Prioritize performance on edge or cost-budgeted cloud setups.
- Architect and deploy scalable, low-cost inference pipelines using vLLM or equivalent engines (2024 stack).
- Optimize for GPU scheduling, high-density Kubernetes deployment, such as:
- QUANTIZATION/POST-TRAINING noise scaling.
- Batching strategies to achieve sub-$0.01 cost targets per HMIS form at 10M+ scale.
- Build future-proof systems:
- Local/cloud-on-premise deployment (targeting EdgeGPU devices like Jetson, limited cloud spots).
- Offline and low-bandwidth sync capabilities.
- Evaluate and commission next-gen hardware (e.g., NVIDIA Blackwell nodes) for balancing:
- 10M-form monthly capacity with cost efficiency.
- Lead technical enablement for new functionality:
- Offline/edge processing support.
- Multilingual support beyond English (e.g. Arabic-Swahili).
- Ensuring compatibility with health data standards (foreign: Dhis2).
- Willingness to:
- Travel to PATH’s rollout countries (DRC, Nigeria, Ghana).
- Overlaps in GMT-1 (Ethiopia/Brazil) and ESA+1 (Pakistan) working hours.
Reasons to use Rodeo
Start with a chat, not a search bar
Grad scheme, placement, apprenticeship? Not sure what you want yet — that's fine. Your agent talks it through with you and turns "I have no idea" into a shortlist.
It searches the market for you
Every day your agent scans the market matching roles against what actually matters to you, not just keywords on a CV.
Only hits
No noise. No "maybe this fits." Just roles with a clear explanation of why they're right — and where to focus when applying.
Requirements
**Required Qualifications and Experience
- Education: B.S/B.A. or M.S. degree in Computer Science, AI, ML, or related quantitative field.
- Experience:
- 7+ years in Machine Learning Engineering, with at least 1–2 years specialising:
- Document AI,
- Computer Vision (OCR, HTR), or
- Vision-Language Models.
- Deep expertise in advanced PyTorch workload anatomy focusing on:
- Transformers pipelines,
- Hugging Face ecosystem (optimization via diffusers, PEFT plugins/Mesh).
- Hands-on production-hosting for vLLM deployments including:
- GPU memory streamlining,
- KV_cache managment for high-throughput models (M10/80GB or equiv).
- 7+ years in Machine Learning Engineering, with at least 1–2 years specialising:
- Technical Competencies:
- Specialization in semantically-aware models e.g.:
- Alignment for OCR-as-a-service (Tabletop with multimodal LLM grounding).
- Solutions for image±table parsing (CombinerNet or LISA-based).
- Proficiency in processing:
- Real-world low-quality images (defocus, background noise).
- Image/tiling strategies to handle > 6MB multi-formats (PDF, PNG).
- Experience spanning:
- OpenCV to OpenMIM(level)
- Format conversion pipelines (CVS/JSON).
- Dockerized cluster orchestration (K8s EBS volumes).
- Custom MPIxPL or Ray options for loop-free async.
- Specialization in semantically-aware models e.g.:
- Cloud & Infrastructure:
- Mastery of:
- EBS-optimized GPUs for inference;
- Budget-compliant spot/bid strike across CentOS/Ubuntu Lite.
- Familiarity with Linode-like local cloud provisioning.
- Mastery of:
Get help with your application
Your very own career expert that helps elevate your application to the next level.
**Technical Tooling
- Python-Pandas-CكوDF (state-of-the-art compute-heavy benchmarking, M2-promonomics).
- LangChain / LangSmith-compatible pipelines (focus on schema augmented context shard allocation).
**Personal Traits
- Flair for cost–accuracy–speed trade-off modeling with low-budget constraints.
- Agnostic OSS champion with commitment to:
- Modular OCR plugins (extensible default baseline models).
- OpenDL and incremental alignment.
- Ability to diagnose production accidents in pre-deployed models using select vision probes.
**Key Qualifiers (High-Demand)
- Experience with:
- Mimido adults’ quality images or contamination filters via diffusion pretraining.
- 30%+ latency reduction for base-engine threading anomalies.
- Custom GPU pipeline (core unloading for DLT-sum inside Precision Pro).
- Multilingual awareness: Training pipeline templates for 6-country language tags.
Why Apply:
- Opportunity to transform digital health systems through scalable HTR/dataflow automation.
- Competitive benefits: salary + flexibility, aligned part-time international commitments (formal trilogy funding scope).
PATH is an equal opportunity employer. Apply via employee portal. Competition limited to workers with PATH ACCES Edinburgh or remote access permissions—valid until further notice.
“It took my CV and asked me questions relevant to understanding what kind of jobs to suggest for me. Suggestions were almost perfect. Jobs were exactly what I’ve been looking for.”
Jessica, London
Skills