Dharana Awasthi

Zugado, India (remote). AI ranking, evaluation and experimentation

• Defined requirements and an API contract for a shared ranking and retrieval service covering eligibility filtering, vector-similarity matching and multi-signal scoring, to eliminate duplicated relevance logic across features, with agreed targets for p95 latency, precision@5 and time-to-first-bid before the v1 launch.
• Drove an experimentation framework covering event tracking, feature flags, A/B testing, config-driven weights and guardrail metrics, removing release dependencies for ranking changes and measuring every change against fill rate.
• Led an evaluation and release-gate process using a labelled benchmark set, precision@k thresholds and LLM-as-judge rubrics to block quality regressions before launch, with a pass/fail gate on every ranking or model change.
• Owned AI governance requirements, including PII masking policies, human-in-the-loop moderation, audit logging and rollback criteria, tracked through false-positive rate, review turnaround and contact-sharing incidents.

YogDharana.life. B2C wellness product

• Launched a 0→1 consumer wellness product (web onboarding platform and live programmes) to improve adherence in lifestyle-condition care, reaching 180 paying customers and $38K revenue with no paid marketing.
• Diagnosed a week-2 retention drop-off through attendance data, surveys and interviews; redesigned onboarding with challenges and 1:1 feedback loops, tripling daily attendance from 30% to 90%.
• Built a referral-led growth engine (workshops, corporate wellness, SEO), driving 55%+ repeat customers and 10%+ premium upgrades.

Genpact. GenAI and data platform, US telecom client

• Led a GenAI ETL copilot on Amazon Bedrock from technical requirements through implementation, delivery and monitoring; it generates and validates pipeline logic, replacing a slow manual build-and-review cycle; delivered data products 3x faster with 70% less manual engineering effort.
• Defined success metrics and a human-in-the-loop validation gate (automated Hive/Athena data and runtime checks, UAT) to solve low trust in AI-generated code, driving adoption across client engineering teams.
• Owned roadmap, budget, QA and launch for 70+ big data and AI initiatives on network and security use cases, generating $600K in annual revenue and standardising delivery on one reusable structure.
• Restructured the delivery operating model, solving capacity and cost constraints and reducing team size by 60% (10 to 4) within six months while maintaining throughput across 70+ initiatives.
• Aligned 10 cross-functional teams across US and India time zones; promoted to Lead in 12 months, recognised by the client VP and awarded top-performer bonuses (FY2023, FY2023–24).

IBM. US insurance client (Fortune 100)

• Built a cloud-native AWS data platform (PySpark, Glue, S3, Lambda, Athena) with event-driven ingestion for insurance policy workflows, leading a 5-person team from requirements to production.
• Designed automated data validation and balancing in Athena and Terraform-based orchestration, giving business teams reconciled policy data and repeatable, version-controlled releases.
• Earned three Manager’s Choice Awards for high-impact delivery and cross-functional leadership.

Infosys. Automotive client

• Worked as Scrum Master: translated business requirements into data products, prioritising the backlog and running sprints for the data engineering team.
• Optimised Spark and Hive SQL pipelines to fix slow production jobs, making them 80% faster and owning end-to-end production monitoring.

IBM. Airline and retail clients

• Automated ETL operations with Python and shell scripting, removing repetitive manual work and saving 500+ engineering hours a year.
• Kept mission-critical airline and retail data platforms at 99.95% uptime through provisioning, engine recovery and DEV/QA/PROD release migrations.

Personal product, built solo

• Took the product from problem discovery and a customer survey to production and monitoring: React, Node/Express, SQLite and Clerk sign-in.
• Built a “Help me start” AI feature behind one model gateway with a second-provider fallback and a local template, inside a 4-second latency budget.
• Added input and output guardrails (PII redaction, crisis-language handling, no shaming or labelling) and a 42-case evaluation gate in CI (pass rate, false blocks, safety).
• Raised answers rated good in human review from 64% to 86% with a prompt rewrite, above an 80% launch bar; designed a 50/50 rollout for an intent-to-treat test.

IE Business Impact Lab with Mastercard

• Designed an agentic AI hospitality solution, including problem framing, proof-of-concept design, scenario planning, product roadmap and revenue and profit model.

MBA final integration project with peeSafe

• Built the European go-to-market strategy for 10 products across Germany, France and the Netherlands.