Real projects. Measurable results. A cross-section of our work across AI, ML, RPA, data science, cloud, and cybersecurity — across industries from healthcare to finance to retail.
Manual claim submission was consuming 8 FTEs and taking 3–5 days per batch — with a 4.2% error rate creating costly rework cycles and delayed reimbursements.
"Yeskay's team didn't just build bots — they redesigned our entire claims workflow. The ROI within 6 months was beyond what we projected."
The billing team was manually extracting patient data, cross-referencing with eligibility systems, completing claim forms, and submitting to 12 different payers — each with slightly different formats and requirements. Errors went undetected until rejection notices arrived days later.
We deployed a UiPath automation with ABBYY IDP to extract structured data from EHR exports, validate against payer-specific rules, auto-correct common errors, and submit claims 24/7. The system handles 12 payer formats and routes exceptions to human reviewers with full context.
New hire onboarding required 34 manual handoffs across HR, IT, and Compliance — taking 3+ weeks and consistently delaying productive start dates.
End-to-end Power Automate bots orchestrating account creation, system access provisioning, compliance training enrollment, and equipment requests — triggered by a single HRMS record creation.
Finance team was spending 6 days each month-end manually matching 80,000+ transactions across banking, POS, and ERP systems — with frequent discrepancies requiring manual investigation.
UiPath bots that automatically pull, normalize, and match transactions across all three systems — flagging only genuine exceptions for human review, with a full audit trail.
Existing rule-based fraud system had a 32% false positive rate — blocking legitimate transactions and frustrating customers — while still missing 8% of actual fraud.
"The false positive reduction alone paid for the engagement in the first quarter. Our customer complaint rate on declined transactions dropped 71%."
Rule-based systems couldn't adapt to new fraud patterns — every new fraud vector required manual rule updates. The bank needed a system that could learn from new attacks automatically while dramatically cutting false positives.
We built an ensemble ML model (gradient boosting + neural network) trained on 3 years of transaction history with behavioral features. Deployed on AWS SageMaker with real-time inference, the model self-updates weekly on new fraud patterns with an automated retraining pipeline.
Engineers were spending 30–40% of their time searching across 200,000+ pages of technical documentation spread across SharePoint, Confluence, and legacy PDFs.
RAG-based AI assistant using GPT-4 and Pinecone vector search — ingesting all documentation sources and enabling natural language Q&A with source citations and version awareness.
Inventory planning relied on spreadsheets and buyer intuition — resulting in 18% overstock carrying costs and frequent stockouts during peak seasons that drove customers to competitors.
"For the first time, our buyers have actual data behind their decisions. The forecast accuracy during Black Friday this year was 94% — our best ever by far."
The company had years of rich sales data across 40,000 SKUs but no systematic way to use it. Buyers were making $180M in annual inventory decisions with Excel models and seasonal gut feel — and getting burned every peak season.
We built a multi-horizon forecasting platform using Prophet + XGBoost ensembles, incorporating promotional calendars, weather, and external demand signals. Paired with a Power BI dashboard giving buyers SKU-level recommendations with uncertainty bands and automated reorder alerts.
High readmission rates were incurring CMS penalties and straining care coordination teams who had no way to prioritize which discharged patients needed follow-up most urgently.
ML model scoring every discharged patient's 30-day readmission risk, integrated directly into the EHR care coordinator workflow with automated follow-up task creation for high-risk patients.
A digital health startup needed to launch their AI-powered diagnostics platform in 3 months — with full HIPAA compliance and the ability to scale from 100 to 100,000 users without re-architecture.
End-to-end AWS infrastructure design and build: VPC with private subnets, ECS Fargate for containerized inference, encrypted RDS for PHI, CloudWatch + GuardDuty for compliance monitoring, and Terraform IaC for repeatable deployments across dev/staging/prod.
Rapid growth had led to unmanaged cloud sprawl — the company was spending $2.8M/year on AWS with no FinOps governance, significant overprovisioning, and zero use of savings plans.
8-week optimization engagement: rightsizing 140+ EC2 instances, implementing spot fleets for batch workloads, committing to Reserved Instances and Savings Plans, and establishing tagging governance and budget alerting.
Security team of 6 analysts was drowning in 40,000+ daily alerts — genuine threats were getting lost in the noise, and analyst burnout was causing turnover.
ML-based alert triage and correlation model trained on 18 months of Splunk logs — automatically scoring and clustering alerts, reducing analyst workload to 1,200 prioritized alerts daily, with automated playbooks for the top 20 alert types.
A large law firm suspected data exfiltration risk from departing employees but had no systematic way to detect unusual document access or export behavior across their DMS and email systems.
User & Entity Behavior Analytics system building individual behavioral baselines and detecting deviations — flagging anomalous document access, bulk downloads, and after-hours activity with risk scores for security review.
Tell us about your challenge. We'll show you how we've solved similar problems — and what we'd build for you.