Aliya Financial Technologies
Aliya builds category-defining software to automate lending end-to-end and transform how banks serve their customers — enabling them to expand credit access without added risk. The platform leverages bank transaction data and advanced AI-powered analytics to enhance decision-making, streamline customer experiences, and ensure regulatory compliance.
As Senior Data Scientist, I apply my deep expertise in complex data analysis, statistical modeling, and machine learning to develop solutions that transform lending. Our flagship product, aSCORE, is a regulatory-compliant, alternative measurement of borrower credit risk and resiliency — validated by a top 10 U.S. bank and reviewed by regulatory examiners.
Aliya's mission is to positively impact the financial well-being of every hardworking American by giving them access to less expensive, more transparent loans and other financial products.
Insight AI
An autonomous AI agent for medical research that connects researchers with vast medical data sources including PubMed, NIH RePORTER, ClinicalTrials.gov, MyGene, and MyVariant. Researchers input their objectives and receive comprehensive analysis — scientific summaries, hypotheses, experimental designs, and target identification — with reliable citations and no AI hallucinations.
- Achieved over 1,000 research objectives in first two weeks of beta
- Recognized by the Llama Index community
- Transforms hours of literature review into minutes of AI-powered discovery
TraffMind AI
An advanced video analytics platform for traffic engineering, smart city infrastructure, and transportation data collection. TraffMind provides AI-powered solutions for traffic counting, bicycle and pedestrian analysis, and law enforcement applications.
In partnership with JAMAR Technologies, we launched the Jamar AI Portal — a platform that takes user-uploaded traffic video and applies advanced AI algorithms to deliver traffic data analysis. Traffic studies shouldn't depend on proprietary cameras or weeks of manual work.
Challenger School — AI-Powered Grading Platform
Designed and built a cloud-native platform that automates K-8 exam grading — from score extraction to AI-assisted grading with human quality assurance.
- Score Extraction: Built an AI vision pipeline that reads scanned exam cover pages, extracts student names and scores, and fuzzy-matches to class rosters — reducing score entry workload by 92%
- Digitized Grading: Replaced paper-based grading of handwritten exams with a digital dual-grader workflow (blinded QA), cutting grading labor by 48% in the pilot
- AI Grading: Deployed parallel AI graders achieving 95.3% accuracy vs. human ground truth at ~$0.01/question, projecting a 66% total workload reduction across K-8
- Fine-Tuning: Built a full fine-tuning pipeline that generates training data from human-graded responses and iteratively improves AI grading models per subject
- Scalable Architecture: Cloud-native platform built to scale to any EdTech workload and designed for easy plug-in of any current or future machine learning / AI models
TP53-Mutant AML
Led a multi-year research program investigating therapeutic vulnerabilities in TP53-mutant acute myeloid leukemia (AML) — one of the most challenging cancers to treat. This work spanned PLK4 targeting, single-cell multiomics, and venetoclax resistance mechanisms.
- PLK4 Targeting: Discovered that PLK4 inhibition triggers irreversible polyploidy and apoptotic cell death selectively in TP53-mutant cells. Contributed to an ongoing clinical trial (NCT04730258).
- Single-Cell Multiomics: Using scDNA+protein sequencing and CyTOF, revealed that TP53-mutant clones undergo monocytic differentiation as a venetoclax resistance mechanism. Built an ML model with >95% accuracy in predicting mutation status.
- ASH 2024 Oral Presentation — highlighted as one of the "Key Advances in AML" by The Practical Hematologist.
EVI1+ AML
Investigated the role of the oncogene EVI1 in hematopoietic stem cell differentiation and leukemia — a body of work spanning my PhD thesis and early postdoctoral fellowship. EVI1-overexpressing AML (driven by chromosome 3q26 rearrangements) is considered incurable and occurs in 8–10% of adult AML.
- Nature Communications (2018): Demonstrated that EVI1 overexpression reprograms hematopoiesis via upregulation of Spi1 (PU.1), the master myeloid regulator. All aged mice developed AML at 90–119 days.
- Leukemic Stem Cell Quiescence: Discovered that MECOM activation promotes chemoresistance by upregulating CDKN1C/P57Kip2 — a novel mechanism explaining treatment failure.
- Funded by NIH/NHLBI F31 and NIH/NCI F32 fellowships.