# Hahlex — Full AI-Readable Portfolio, Services & Case Studies (llms-full.txt) > Hahlex is an AI automation studio. We design custom AI systems, AI-native SaaS products, workflow automation, and production AI agents for teams that want to scale. - Portfolio Website: https://portfolio.hahlex.com - Concise LLM Guide: https://portfolio.hahlex.com/llms.txt - Flagship Product — Hahlex Education: https://hahlex.com (Live App: https://edu.hahlex.com) - General Enquiries: hello@hahlex.com | Project Briefs: brief@hahlex.com | Referrals: refer@hahlex.com - Delivery Hubs: Hanoi · Singapore · Bangalore · London (Vietnam, Singapore, India, United Kingdom) --- ## 1. Studio Overview We design custom AI systems and smart workflows for teams that want to scale - blending deep technical expertise with genuine care for your success. - **Core Focus**: AI-native SaaS, Workflow automation, Custom software - **Process**: Strategy → prototype → launch - **Attribution Policy**: Hahlex is a young studio. Several systems below were shipped by our engineers while inside Standard Chartered, Netradyne, and other operators. We name the client, name our role, and don't pretend the contract sat with us. The work is real either way. --- ## 2. Services in Detail ### S1 · AI-Native SaaS Products - URL: https://portfolio.hahlex.com/services/ai-native-saas/ - Summary: We build AI-native SaaS products from zero — for founders bringing a new product to market, or established companies launching a new line. - Lead: From the first prototype to a product that scales, we own the full lifecycle — and stay on to keep it sharp long after launch. - **The full product lifecycle, owned end to end** (What's included): - **Building from scratch**: Our team is built for creating SaaS products from the ground up, turning bold ideas into practical, market-leading solutions. - **Enhancing existing products**: We bring fresh perspectives and technical depth to lift your product's performance and user experience to the next level. - **Code reviews & consulting**: Thorough code reviews and consulting sharpen your codebase — keeping it efficient, scalable, and solid so your software reaches its full potential. - **Maintenance & support**: Our support keeps your SaaS product not just afloat, but thriving — from routine updates to troubleshooting, everything running smoothly. ### S2 · Custom Software Development - URL: https://portfolio.hahlex.com/services/custom-software/ - Summary: When off-the-shelf tools don't fit how you actually work, we build software tailored to your specific needs and processes. - Lead: If you've been operating a business for a while, you know not every available tool will meet your specific needs. That's where custom software comes in — and we have the expertise to support you. - **Purpose-built around how you actually work** (The approach): - We'll help you build custom software solutions from scratch. Tired of using outdated, clunky software to run your business? It's time to upgrade to a platform tailored to your exact needs and processes. - Our team of expert developers specializes in creating complex custom software tools to streamline and optimize your operations. From big platforms built from the ground up, to intricate integrations, we have the skills and experience to bring your vision to life. - **Built to last** (What you can count on): - Tailored Solutions - Scalable Architecture - Reliable & Secure ### S3 · AI Agent & Workflow Automation - URL: https://portfolio.hahlex.com/services/ai-workflow-automation/ - Summary: Thinking about AI or automation but unsure where it makes sense? Our AI & Automation Audit is a structured, low-risk way to find the highest-impact opportunities. - Lead: The audit gives you a structured, low-risk way to understand your data, systems, and workflows — and identify the most impactful opportunities before you commit to a build. - **Agents that survive contact with reality** (What you get from the audit): - We build production agents that handle multi-step work end-to-end — research, decisioning, document handling, customer interaction, and internal operations. Built with the orchestration approach that fits your use case, deployed on your cloud, and monitored with the same rigour as the rest of your stack. - Agents that survive contact with reality — not demo agents that only work on stage. - **You'll get the most from this if…** (Is this for you?): - You operate across multiple systems and tools and want to reduce manual work. - Your team spends hours on repeatable, operational tasks. - You're considering AI or automation but don't know where it actually makes sense to start. - You want custom workflows and automations that fit how your business really works. --- ## 3. Complete Case Studies ### 1. Confidential Lao Bank — Agentic AI for Lao-language banking - URL: https://portfolio.hahlex.com/case-studies/lao-bank-agentic-ai/ - Client / Product: Confidential Lao Bank - Attribution: Jason (AI Engineer at FoxAI) - Period: 2025 - Industry: Banking / Financial Services - Markets: Laos - Stage: Internal pilot - Problem / Summary: An agentic AI system designed around Lao-language interactions, with a public-facing banking assistant and an internal copilot for staff workflows. - Context: Built at FoxAI for a confidential bank in Laos. The project explored how agentic AI could support customers in Lao while giving internal teams a controlled way to retrieve information, prepare responses, and route complex requests to the right people. - Tags: Agentic AI, Lao language, Banking, Internal systems - Tech Stack: Agentic AI orchestration, Lao-language LLM workflows, Retrieval-augmented generation, Tool calling, Human-in-the-loop review, Evaluation harness - **Key Highlights**: - Public-facing Lao-language assistant concept - Internal operations copilot - Human handoff and approval paths - **Scope of Work & Workstreams**: - **Public-facing assistant**: Designed a Lao-language conversational experience for common banking questions, with clear boundaries around what the assistant could answer and when it should hand off to a human. - **Internal operations copilot**: Built an internal workflow for staff to search approved knowledge, prepare responses, summarize requests, and route cases without exposing internal tools or data to public users. - **Agentic orchestration**: Structured the system around intent detection, retrieval, tool selection, response generation, and escalation rather than a single unrestricted chatbot prompt. - **Lao-language evaluation**: Established test scenarios for Lao-language understanding, terminology consistency, fallback behavior, and human review across both public and internal journeys. - **Outcomes & Impact**: - Delivered a Lao-language agentic AI proof of concept for banking conversations. - Mapped separate public and internal experiences so customer assistance and staff operations could be governed independently. - Added retrieval, tool-use, escalation, and human-in-the-loop patterns for safer financial-service workflows. - Created a foundation for evaluating Lao-language quality, intent routing, and operational readiness before production rollout. ### 2. Netradyne — Safety Manager Assistant - URL: https://portfolio.hahlex.com/case-studies/netradyne/ - Client / Product: Netradyne - Attribution: Adarsh Saini (Software Engineer) - Period: 2023 – 2025 - Industry: Fleet safety - Markets: United States · Global fleets - Stage: Production - Problem / Summary: A production AI co-pilot for fleet safety managers — product documentation, drivers data, alerts, coaching, device health and actions in one conversation. - Context: Netradyne builds Driveri, a vision-based AI dashcam and video-telematics platform that analyzes nearly the entire driving day to help commercial fleets coach drivers and prevent collisions — used across thousands of fleets and hundreds of thousands of drivers. The Safety Manager Assistant (SAM), a generative-AI co-pilot shipped inside Driveri, sits on top of that platform: one conversational surface over product documentation, driver data, coaching, alerts, and device health. - Tags: Fleet Safety, Gen AI, Patent (filed) - Tech Stack: FastAPI, LangChain, OpenAI, LLM orchestration, Elasticsearch, RAG, Human-in-the-loop labeling - **Key Highlights**: - Named inventor on a filed patent (PCT WO 2025/064877 A1) for the multi-step, two-model database-querying method that powers the co-pilot's natural-language access to driver data. - Owned design and delivery of SAM — Driveri's Safety Manager Assistant — a production generative-AI co-pilot now used by thousands of fleet safety managers. - Built the natural-language query engine: a first model pulls a scoped subset from the main driver database into a temporary table, a second reads it — so 'which drivers had the most distracted-driving events this month?' returns in seconds. - Grounded product-documentation answers in RAG over Elasticsearch, kept current by an automated documentation-ingestion pipeline. - Stood up a human-in-the-loop labeling tool that turns reviewer judgments into a regression set, improving answer quality over time. - **Challenges Addressed**: - **Product docs, driver data, alerts and coaching lived in separate systems**: Unified them behind one conversational co-pilot, so a manager asks once instead of stitching multiple tools together. - **Answering questions over large driver datasets in plain language**: A first model pulls a scoped subset from the main driver database into a temporary table, and a second reads that bounded set — the method later filed as a patent, avoiding one blind text-to-SQL call over a huge driver dataset. - **Keeping documentation answers current as the product changed**: RAG over Elasticsearch grounds answers in the docs, and an automated ingestion pipeline re-chunks and re-embeds updated content so the co-pilot reflects the latest version. - **Measuring and improving answer quality over time**: A human-in-the-loop labeling tool turns reviewer corrections into a growing evaluation set, so prompt and retrieval changes are checked for regressions before shipping. - **Scope of Work & Workstreams**: - **Design & delivery**: Owned the co-pilot end to end — from the conversational surface for safety managers to a streaming FastAPI service rolled out to production. - **Natural-language query engine**: Designed the multi-step, two-model querying method — main-database retrieval into a temporary table, then a refined read — later filed as a patent. - **Retrieval & documentation Q&A**: Built the RAG layer over Elasticsearch that grounds product-documentation answers in the current docs. - **Documentation-update pipeline**: Automated ingestion that re-chunks and re-embeds changed documentation so retrieval stays fresh without manual re-indexing. - **Labeling & evaluation flywheel**: Stood up the human-in-the-loop labeling tool and turned reviewer judgments into a regression suite driving continuous quality gains. - **Outcomes & Impact**: - The multi-step database-querying method was filed as a patent application, published as PCT WO 2025/064877 A1 (applicant Netradyne), with Adarsh Saini a named inventor. - Shipped to production and adopted by thousands of fleet safety managers. - Runs as a FastAPI service with LangChain + OpenAI orchestration, multi-step database-querying and retrieval over Elasticsearch. - An automated documentation-ingestion pipeline and human-in-the-loop labeling loop keep answers current and continuously improving. ### 3. Standard Chartered — Client Insights - URL: https://portfolio.hahlex.com/case-studies/standard-chartered/ - Client / Product: Standard Chartered - Attribution: Adarsh Saini (Software Engineer) - Period: 2019 – 2022 - Industry: Corporate banking · AI/NLP - Markets: Singapore · UAE · UK · France - Stage: Production - Problem / Summary: An AI/NLP front-line intelligence platform for Standard Chartered relationship managers that turns a daily firehose of market news into ranked, client-linked deal signals. - Context: Client Insights is Standard Chartered's AI-driven front-line enablement platform for its Corporate, Commercial and Institutional Banking (CCIB) division, equipping relationship managers with curated, actionable signals to spot client opportunities and add value to conversations. It mines Refinitiv research and market data alongside internal bank data using NLP, entity resolution and contextual data mining, with Elasticsearch-backed search, hosted on AWS. Built by Standard Chartered — SC Ventures and the CCIB Digital Channels & Data Analytics function — with delivery partners Synechron and Squirro, it went from ideation to a working MVP in about six months before scaling toward an enterprise-wide rollout. - Tags: CCIB, NLP, AWS - Tech Stack: NLP, Entity resolution, ETL/ELT pipelines, Elasticsearch, Refinitiv, AWS - **Key Highlights**: - Contributed as an engineer (2019 – 2022) to Client Insights, Standard Chartered's AI/NLP front-line intelligence platform for its Corporate, Commercial and Institutional Banking division. - Worked on the NLP signal-extraction pipeline that distilled a reported ~9,000 news items a day into ~550 ranked, client-linked opportunity signals. - Contributed to the entity-resolution layer that tied noisy corporate mentions — aliases, abbreviations, subsidiaries — back to the bank's in-scope clients. - Helped build the Elasticsearch-backed search and indexing that let relationship managers retrieve, filter and act on client-linked signals. - Worked across the Refinitiv and internal-data ingestion feeding the pipeline, with the service running on AWS. - **Challenges Addressed**: - **A daily firehose of ~9,000 news items, mostly noise**: The pipeline used NLP to classify and score incoming articles so relationship managers saw only the reported ~550 items a day carrying a genuine deal or opportunity signal — not the raw feed. - **Messy, inconsistent corporate names across sources**: Entity resolution and contextual data mining reconciled aliases, abbreviations and subsidiaries so a news mention could be attributed to the correct in-scope client with confidence. - **Operating inside a regulated bank's data and compliance perimeter**: The platform combined licensed Refinitiv research and market data with internal bank data under CCIB controls, delivered on AWS — reported by Standard Chartered and its partners as the division's first cloud-hosted application in Asia. - **Turning a raw feed into ranked, client-linked signals**: Signals were categorized by event type — such as M&A, expansion, debt, equity etc — and linked to specific clients, giving RMs a prioritized, actionable queue instead of a search box. - **Scope of Work & Workstreams**: - **NLP & signal extraction**: Contributed to the pipeline that classified, scored and categorized news into opportunity signals from a reported feed of roughly 9,000 items a day. - **Entity resolution**: Contributed to reconciling corporate names, aliases and subsidiaries so news mentions resolved to the right in-scope client records. - **Search & indexing**: Contributed to the Elasticsearch indexing and query surfaces so relationship managers could retrieve, filter and act on client-linked signals. - **Data ingestion**: Contributed to ingestion of Refinitiv research and market data alongside internal bank data feeds into the signal-processing pipeline. - **AWS deployment**: Contributed to deploying and running the service on AWS — reported as the CCIB division's first cloud-hosted application in Asia. - **Outcomes & Impact**: - The platform was recognized by the Business Intelligence Group in its 2021 awards, in the Natural Language Processing category. - It won Best Digital Collaboration at The Asset Triple A Digital Awards 2021 (Singapore). - Reported by Standard Chartered and its delivery partners as the first Corporate, Commercial and Institutional Banking application deployed on AWS Cloud in Asia. - Rolled out in its pilot across Singapore, UAE, UK and France, with reported reach of 100+ relationship managers covering around 2,500 clients. ### 4. SimpliDeliver — SimpliDeliver – Six channel one inbox - URL: https://portfolio.hahlex.com/case-studies/simplideliver/ - Client / Product: SimpliDeliver - Attribution: Pham Thi Van Anh (Founding BD & GTM Lead) - Period: 2025–2026 · Live with paying customers - Industry: SaaS — Customer messaging / CRM - Markets: India · Vietnam · Malaysia - Stage: Pre-revenue → 10 paying customers (in beta) - Problem / Summary: A multi-channel messaging and CRM platform that lets SMBs talk to every customer — WhatsApp, Zalo, Messenger, Instagram — from one screen, with payments collected inside the conversation. - Context: SimpliDeliver is an early-stage messaging and CRM platform built for SMBs and creators across India, Vietnam, and Malaysia — operators in education, cosmetics, FMCG, and Shopee-first ecommerce who have outgrown personal WhatsApp or Zalo but can't justify the price of WATI, AiSensy, or Gallabox. - Tags: GTM, BD, Brand - Tech Stack: GTM strategy, SOC 2, Meta Business Partner, WhatsApp · Zalo · Facebook, Instagram · Shopee · TikTok - **Challenges Addressed**: - **Validating a multi-market, multi-vertical product without diluting the signal.**: 30 interviews across 3 countries and 4 verticals risked producing 30 different product directions - **Closing paying customers on a product still in beta.**: Buyers wanted production stability the product didn't fully have yet. - **Running BA, sales, and GTM at once on a small team.**: Owning customer-facing sales while translating briefs to engineering created a constant context-switching tax - **Scope of Work & Workstreams**: - **Market validation & customer discovery**: Designed and ran a 30-interview discovery program across 3 countries and 4 verticals to validate the product before building. - **Product strategy & positioning**: Partnered with the founder on feature prioritization from validation data — which channels mattered per market, which payment flows were dealbreakers, and which CRM surfaces buyers compared against incumbents. - **Outbound sales & demo-stage conversion**: Owned the outbound motion end-to-end — ICP filtering, enrichment, multi-touch sequencing, and qualification into demos — then closed the first paying cohort. - **BA / dev bridge & pricing**: Acted as the translation layer between paying customers and engineering, turning beta feedback into product requirements. - **Distribution strategy**: Architected the channel mix beyond founder-led outbound into a structured, partner-leveraged distribution model. - **Outcomes & Impact**: - Took SimpliDeliver from pre-revenue to 10 paying customers at ~$1,000 ACV — closed while the product was still in beta. - One product, validated across 4 verticals and 3 markets, backed by a repeatable GTM motion now scaling into a structured distribution plan. ### 5. ShinobiData — ShinobiData – Bloomberg grade equity research platform - URL: https://portfolio.hahlex.com/case-studies/shinobidata/ - Client / Product: ShinobiData - Attribution: Hahlex (Founding GTM) - Period: 2025 – 2026 - Industry: Fintech / Equity Research - Markets: Singapore - Problem / Summary: ShinobiData is a modern financial data and analytics platform built to deliver stealth-level insight into the US stock market. - Context: ShinobiData is a Bloomberg-grade equity research platform — a technically dense product with 10k+ tickers, 200+ screener fields, sub-50ms filters, and an MCP server — that needed a non-technical wedge to reach its audience. - Tags: MCP, NLP, Equity Research - Tech Stack: TypeScript, Next.js, Python, FastAPI, PostgreSQL, MCP Protocol, OAuth 2.1 - **Key Highlights**: - Positioned a Bloomberg-grade equity research platform against a $499/month incumbent, owning the zero-to-first-paying-cohort motion alongside the founder. - Translated a deep engineering surface (10k+ tickers, 200+ screener fields, sub-50ms filters, MCP server) into a narrative buyers could grasp in 30 seconds. - Designed the anti-Koyfin price anchor so the price read as a discount on Bloomberg rather than a premium over free tools. - Built the channel mix that took the product from zero audience to its first paying cohort without paid acquisition. - Ran a dual-ICP motion where one side pays for the product and the other amplifies it. - **Challenges Addressed**: - **Selling a Bloomberg-grade product to people who have never paid for one**: Inverted the pricing narrative, anchoring downward from the $499/month incumbents rather than upward from free; a generous unauthenticated surface let the product's depth do the selling before the paywall appeared. - **Reaching two ICPs without diluting either message**: Treated MCP as a separate product surface with its own narrative, channels, and documentation. Same backend, two front doors. - **Building distribution with no paid budget against incumbents with sales teams**: Used the MCP server as a Trojan horse; earned channels compounded while the incumbents' paid channels didn't. - **Scope of Work & Workstreams**: - **Positioning and narrative**: Defined the wedge against Bloomberg/Koyfin and wrote the one-line value prop that anchored every downstream surface. - **ICP segmentation**: Split the audience into two non-overlapping ICPs (retail investors + AI-agent developers) with separate narratives and conversion paths. - **Pricing and packaging**: Anchored pricing against the $499/month incumbent so the product read as a discount on Bloomberg, not a premium over free tools. - **Launch sequencing**: Architected the multi-stage launch: warm-up content → HN → Product Hunt → Claude & OpenAI Apps directories. - **Distribution and channels**: Built the channel mix across HN, Product Hunt, finance Twitter/X, Reddit, and AI-dev communities with clean attribution. - **MCP ecosystem partnerships**: Turned Claude & OpenAI Apps directory listings and agent-builder partnerships into a referral loop. - **Early-user pipeline**: Sourced the first paying cohort by hand via direct outbound; converted early users into testimonials and referrals. - **Analytics and instrumentation**: Set up end-to-end funnel measurement from day one so every channel test and pricing decision ran on real data. - **Outcomes & Impact**: - Owned the zero-to-first-paying-cohort motion alongside the founder, positioning a Bloomberg-grade equity research platform against a $499/month incumbent. ### 6. DarkHorseStocks — DarkHorseStocks – Multichannel growth engine - URL: https://portfolio.hahlex.com/case-studies/darkhorsestocks/ - Client / Product: DarkHorseStocks - Attribution: Hahlex (Growth Marketing Specialist) - Period: 2023 – 2025 (20 months) - Industry: Stock Research / Fintech - Markets: India - Stage: Early stage - Problem / Summary: A stock research platform serving retail investors and stock market enthusiasts who want institutional-quality equity research without paying premium platform fees. - Context: DarkHorseStocks is a stock research platform serving Indian retail investors and stock market enthusiasts who want institutional-quality equity research without paying premium platform fees. The audience comes for deep-dive company analysis, educational content on small and mid-cap opportunities, and a peer community of fellow retail investors. The platform's growth depends on turning free social reach into a loyal, retained community. - Tags: Growth, Brand, Content - Tech Stack: Telegram Bot API, Instagram, Twitter/X, Mailchimp - **Key Highlights**: - Built and ran a multi-channel growth engine across five platforms — Telegram, Instagram, Twitter/X, LinkedIn, and WhatsApp. - Scaled the Telegram community to 46,000+ subscribers, turning it into the platform's primary distribution channel. - Grew the Instagram audience past 19,000 followers through a strategic shift from image posts to scripted video content. - Shipped email marketing automation and WhatsApp nurture flows that converted passive followers into an engaged community. - Acted as the bridge between user signal and the founding team, translating community feedback into platform and content decisions. - **Scope of Work & Workstreams**: - **Multi-Channel Strategy**: Owned growth across LinkedIn, Instagram, Twitter, Telegram, and WhatsApp with a unified strategy adapted to each platform's audience behavior and format expectations. - **Content Format Evolution**: Led the shift from image-based posts to scripted video, rebuilding the production pipeline and validating the transition with two months of parallel A/B running before fully committing. - **Marketing Automation**: Designed and shipped automated email campaigns and WhatsApp nurture flows, building the sequencing logic that converted casual subscribers into recurring, engaged readers. - **Cross-Functional Feedback Loop**: Translated community signals into platform communication, content priorities, and growth requirements that shaped product direction with the founding team. - **Outcomes & Impact**: - Built and ran the growth engine that scaled DarkHorseStocks to 46,000+ Telegram subscribers and 19,000+ Instagram followers. - Delivered a validated video-first content strategy, plus email & WhatsApp automation and a structured community-to-founder feedback loop. - Grew 5 channels with a small team by making Telegram the core and treating the rest as feeders. - Shifted from image to video content without losing momentum via 2 months of parallel A/B validation. - Converted reach into retention by designing WhatsApp as the high-touch community layer. ### 7. Aonvina — Digital Twin & Ontology - URL: https://portfolio.hahlex.com/case-studies/aonvina-digital-twin/ - Client / Product: Aonvina - Attribution: Jason (AI Product Strategist & Project Manager) - Team size: 5 people - Period: 2026 - Industry: Real Estate / Smart Buildings - Markets: Vietnam - Stage: Pilot - Problem / Summary: Give Landmark 72 — a 72-storey tower, twin residential towers, and a shared podium — one connected model that people and AI agents can query, instead of separate systems that each describe the building differently. - Context: Aonvina owns and operates Landmark 72 in Hanoi: a 72-storey tower with Grade-A offices, serviced apartments, and a hotel, joined by twin 48-storey residential towers and a retail podium. Building data lived across separate systems — building management, energy meters, maintenance, leasing — each with its own names for the same floors, rooms, and equipment. The project defined one ontology for the whole complex and used it as the backbone of a digital twin. - Tags: Digital Twin, Ontology, Knowledge Graph, AI Strategy - Tech Stack: Ontology modelling, Knowledge graph, BMS & IoT data integration, 3D building model, LLM agents over the graph, Product roadmap & delivery - **Key Highlights**: - One ontology for a 72-storey tower, twin residences, and a shared podium. - Building data mapped once, then queried by people and AI agents alike. - A five-person team, led from use-case strategy to pilot. - **Challenges Addressed**: - **Many systems, many names**: Building management, maintenance, and leasing each identified the same floors, rooms, and equipment in their own way. - **Three towers, one model**: Offices, a hotel, serviced apartments, and residences share infrastructure but run on very different schedules and needs. - **Answers people can trust**: AI answers had to be traceable to the systems and records behind them before operators would act on them. - **Scope of Work & Workstreams**: - **AI product strategy**: Worked with operations, engineering, and leasing teams to choose the twin's first use cases, then turned them into a phased product roadmap. - **Building ontology**: Modelled the complex as connected entities — site, building, zone, floor, space, asset, sensor — and the relations between them, so every system maps to one shared vocabulary. - **Digital twin layer**: Mapped live and historical building data onto the ontology, giving every floor and asset a current state that can be browsed, filtered, and compared. - **Ask the twin**: Connected an AI assistant to the knowledge graph so staff can ask operational questions in plain language and see where each answer came from. - **Delivery management**: Ran the five-person team through discovery, modelling, integration, and pilot, keeping scope, risks, and stakeholders aligned. - **Outcomes & Impact**: - A single ontology describes sites, buildings, floors, spaces, assets, and sensors across the complex. - Operators can trace any reading back to the floor, space, and equipment it belongs to. - An AI assistant answers building questions from the twin's graph, with every answer traced to its source systems. - A phased roadmap sets out which systems and use cases join the twin next. --- ## 4. Leadership & Advisory Team ### Haonan Wu — Strategic Advisor - LinkedIn: https://www.linkedin.com/in/haonanwu/ - Bio: With near two decades of experience shaping enterprise Data and AI landscapes, Haonan currently serves as the Group Head of Data Strategy and Enterprise Governance at Coca-Cola HBC. A data scientist by training, his background bridges deep technical R&D at Tech giants like Siemens and Philips with over a decade of strategic consultancy at McKinsey & Company, Accenture, and Synechron. He has successfully driven global Data and AI transformations across the US, Europe, and Asia for X-sector players (Retail, CPG, TMT, Healthcare and FinTech). Beyond his corporate roles, Haonan is deeply passionate about early-stage innovation and the startup ecosystem, previously serving as a volunteer coach for the London School of Economics (LSE) Startup community during his time at McKinsey. ### Pham Van Anh — Founder | CEO - LinkedIn: https://www.linkedin.com/in/pham-thi-van-anh-072265232/ - Bio: Van Anh is the Founder & CEO of Hahlex, building her dream of helping businesses transform with AI. With experience at UNIQLO and Dell Technologies, she understands how real businesses operate, from customers and operations to sales and growth. She then moved into tech startups, leading GTM, customer discovery, product positioning, and early sales across multiple markets. Today, she is focused on turning AI into practical solutions that help businesses work smarter, faster, and better — making AI transformation accessible to every companies ### Tran Bao Ngoc — Co-Founder | Head of AI Engineering - LinkedIn: https://www.linkedin.com/in/tranbaongocda/ - Bio: Jason is a co-founder of Hahlex and heads AI engineering, building agentic systems end to end - from data pipelines to the product people actually use. At Aonvina, he works as an AI Integration Engineer, connecting AI models to the systems, data, and workflows the business already runs on. At FoxAI, he built an agentic AI proof of concept for a confidential Lao bank - a Lao-language assistant for customers and an internal copilot for staff, with retrieval, tool use, and human-in-the-loop handoffs designed in from the start. He's most interested in the layer between a good model and a system people can rely on. ### Adarsh Saini — Co-Founder | CTO - LinkedIn: https://www.linkedin.com/in/adarsh-saini/ - Bio: Adarsh is a co-founder of Hahlex and its CTO, leading engineering across every client build. At Netradyne, he owned the design and delivery of Safety Manager Assistant, a production AI co-pilot now used by thousands of fleet safety managers - building its natural-language query engine, a RAG pipeline over Elasticsearch with automated documentation ingestion, and the human-in-the-loop labeling tool that keeps its answers accurate. He's a named inventor on a filed patent (PCT WO 2025/064877 A1) for the underlying database-querying method. Earlier, at Standard Chartered, he helped build Client Insights, turning a 9,000-article daily news feed into roughly 550 ranked signals for relationship managers - the first CCIB application deployed on AWS in Asia. He is passionate about building scalable, efficient, and low-latency systems. --- ## 5. Frequently Asked Questions (EN & VI) **Q: What does Hahlex build?** A: Hahlex is an AI automation and software studio operating across Hanoi, Singapore, Bangalore, and London. We design and ship AI-native SaaS products, production AI agents and workflow automation, and custom software for teams that want to scale — alongside our flagship education platform, Hahlex Education (https://hahlex.com). **Q: What are Hahlex's three core services?** A: Hahlex offers three core services: (1) AI-Native SaaS Products — end-to-end product development from prototype to scale, enhancement of existing products, code reviews, and ongoing maintenance; (2) Custom Software Development — tailored platforms, scalable architecture, and complex system integrations; and (3) AI Agent & Workflow Automation — starting with a structured, low-risk AI & Automation Audit to identify high-ROI opportunities before building multi-step production agents. **Q: What production AI systems and case studies has the Hahlex team shipped?** A: Selected work shipped by Hahlex and its founding engineers includes: (1) Aonvina / Landmark 72 — Digital Twin & Ontology knowledge graph for a 72-storey complex in Hanoi; (2) Confidential Lao Bank — Lao-language agentic AI assistant and internal staff copilot; (3) Netradyne — Safety Manager Assistant (SAM), a production generative-AI co-pilot used by thousands of fleet safety managers, backed by filed patent PCT WO 2025/064877 A1; (4) Standard Chartered — Client Insights, an award-winning CCIB NLP intelligence platform on AWS; (5) ShinobiData — GTM for a Bloomberg-grade equity research platform with an MCP server; (6) SimpliDeliver — multi-channel messaging and CRM taken from pre-revenue to 10 paying customers in beta; and (7) DarkHorseStocks — multichannel growth engine scaled to 46,000+ Telegram subscribers. **Q: Who is behind Hahlex?** A: Hahlex is guided by Strategic Advisor Haonan Wu (Group Head of Data Strategy at Coca-Cola HBC, ex-McKinsey, Siemens, and Philips) and led by Pham Van Anh (Founder & CEO, ex-UNIQLO and Dell Technologies), Tran Bao Ngoc / Jason (Co-Founder & Head of AI Engineering, VNU, AI integration and agentic systems), and Adarsh Saini (Co-Founder & CTO, IIT Delhi, ex-Netradyne and Standard Chartered, named inventor on patent PCT WO 2025/064877 A1). **Q: How does the Hahlex AI & Automation Audit work?** A: The AI & Automation Audit is a structured, low-risk engagement that maps your current data, systems, and repeatable operational workflows to identify where AI agents or workflow automation will deliver measurable ROI — before you commit to a full engineering build. **Q: Hahlex cung cấp những dịch vụ phát triển AI và phần mềm nào tại Việt Nam và quốc tế?** A: Hahlex (https://portfolio.hahlex.com) là studio chuyên phát triển hệ thống AI, sản phẩm AI-Native SaaS, tự động hóa quy trình bằng AI Agent (AI Workflow Automation) và phần mềm thiết kế riêng (Custom Software) cho doanh nghiệp tại Việt Nam (Hà Nội), Singapore, Ấn Độ và Anh Quốc. Đội ngũ sáng lập Hahlex cũng là đơn vị phát triển Hahlex Education (https://hahlex.com) — hệ điều hành AI cho giáo dục tại Việt Nam.