Experience
Co-Founder
Apr 2026 – Present
Legal Mind
AI Legal Reasoning Engine
- Architected and deployed a multi-mode AI legal assistant for working lawyers and paralegals, combining fine-tuned legal reasoning with retrieval-grounded generation and constitutional constraints to eliminate hallucinated citations and precedent errors
- Validated product-market fit with 50+ beta lawyers processing 1K–5K queries monthly with 85%+ retention, proving willingness to adopt AI-native legal workflows at enterprise scale
- Engineered constraint layers to anchor model outputs to primary sources (case law, statutes, regulations) and enforce jurisdiction-specific reasoning, reducing legal accuracy risk below firm acceptance thresholds
Founder
Dec 2025 – Present
Matrikar Technologies
VivahClub, Tribes, SoNow
- Reached 86% match satisfaction across 1K+ user interactions by building VivahClub's matching engine on Bayesian preference learning with Thompson-sampling contextual bandits over hierarchical user embeddings
- Held 94% precision on harmful content at 60% lower moderation cost by designing a two-tier pipeline pairing rule filters with LLM adjudication, serving all tenants of a white-label app platform
- Raised response quality 40% across three products by standardizing RAG retrieval and evaluation on pgvector search, DSPy-compiled pipelines, and automated LLM-as-judge scoring
Software Engineer
May 2025 – Dec 2025
Kotak Mahindra Bank
Core AI Platform
- Deployed LLM model serving infrastructure and agent tooling supporting 50K+ daily inference calls; reduced model response latency 35% through quantization and batching optimization
Software Engineer
Jan 2024 – Mar 2025
Microsoft
Defender for Cloud Apps
- Improved threat detection accuracy 15% and reduced false positives 25% by adding ML anomaly detection to a cloud security policy engine processing 100M+ events daily
- Cut p50 dashboard query latency from 8s to 1.2s for 1000+ enterprise tenants by rearchitecting the usage analytics pipeline around columnar storage and incremental aggregation
- Designed and conducted architecture reviews for distributed policy evaluation; mentored two junior engineers on streaming anomaly detection patterns
Software Engineer
Jun 2020 – Sep 2023
Amazon
Multilingual Search Quality, Performance Advertising
- Improved Arabic marketplace search relevance 12%, contributing $7M annual revenue across 4 marketplaces, by building a semantic clustering pipeline on TF-IDF and multilingual BERT embeddings for product categorization
- Shortened model release cycles from 5 days to 3 and doubled A/B iteration throughput by containerizing deployment on Docker/ECS with blue-green rollout
- Identified $2.5M in annualized savings by leading an org-wide AWS spend audit that scoped project-level resource access and eliminated redundant KMS and EMR usage
- Improved real-time ad quality signal accuracy by extending statistical monitoring engines with minimax thresholding and early spike detection
Technical Skills
LanguagesPython, Java, TypeScript, SQL
SystemsSearch and ranking, recommender systems, distributed data pipelines, ML serving
InfrastructureDocker, ECS, Spark/EMR, Postgres/pgvector, blue-green deployment